AI News Collector

The AI state of the art is shifting rapidly from simple chat interfaces to autonomous agents capable of planning, executing, and refining complex workflows. In this new landscape, the ability to ground these intelligent agents in your enterprise data is key to unlocking true business value. Google Cloud is at the forefront of this shift, empowering you to build robust, data-driven applications quickly and accurately. Last month, Google announced Antigravity, an AI-first integrated development environment (IDE). And now, you can now give the AI agents you build in Antigravity direct, secure access to the trusted data infrastructure that powers your organization, turning abstract reasoning into concrete, data-aware action. With Model Context Protocol (MCP) servers powered by MCP Toolbox for Databases now available within Antigravity, you can securely connect your AI agents to services like AlloyDB for PostgreSQL, BigQuery, Spanner, Cloud SQL, Looker and others within Google’s Data Cloud, all within your development workflow. Why use MCP in Antigravity? We designed Antigravity to keep you in the flow, but the power of an AI agent is limited by what it "knows." To build truly useful applications, your agent needs to understand your data. MCP acts as the universal translator. You can think of it like a USB-C port for AI. It allows the LLMs in your IDE to plug into your data sources in a standardized way. By integrating pre-built MCP servers directly into Antigravity, you don’t need to perform any manual configuration. Your agents can now converse directly with your databases, helping you build and iterate faster without ever leaving the IDE. Getting started with MCP servers In Antigravity, connecting an agent to your data is a UI-driven experience, eliminating the challenges we’ve all faced when wrestling with complex configuration files just to get a database connection running. Here’s how to get up and running. 1. Discover and launch You can find MCP servers for Google Cloud in the Antigravity MCP Store. Search for the service you need, such as "AlloyDB for PostgreSQL" or "BigQuery," and click on Install to start the setup process. Launching the Antigravity MCP store 2. Configure your connection Antigravity presents a form where you can add your service details such as Project ID and region. You can also enter your password or have Antigravity use your Identity and Access Management (IAM) credentials for additional security. These are stored securely, so your agent can access the tools it needs without exposing raw secrets in your chat window. Installing the AlloyDB for PostgreSQL MCP Server See your agents in action Once connected to Antigravity, your agent gains a suite of "tools" (executable functions) that it can use to assist you, and help transform your development and observability experience across different services. Let’s take a look at a couple of common scenarios.  Streamlining database tasks with AlloyDB for PostgreSQL When building against a relational database like PostgreSQL, you may spend time switching between your IDE and a SQL client to check schema names or test queries. With the AlloyDB MCP server, your agent handles that context and gains the ability to perform database administration and generate high-quality SQL code you can include in your apps — all within the Antigravity interface. For example: Schema exploration: The agent can use list_tables and get_table_schema to read your database structure and explain relationships to you instantly. Query development: Ask the agent to "Write a query to find the top 10 users," and it can use execute_sql to run it and verify the results immediately. Optimization: Before you commit code, use the agent to run get_query_plan to ensure your logic is performant. Antigravity agent using the MCP tools Unlocking analytics with BigQuery For data-heavy applications, your agent can act as a helpful data analyst. Leveraging the BigQuery MCP server, it can, for example: Forecast: Use forecast to predict future trends based on historical data. Search the catalog: Use search_catalog to discover and manage data assets. Augmented analytics: Use analyze_contribution to understand the impact of different factors on data metrics. Building on truth with Looker Looker acts as your single source of truth for business metrics. Looker’s MCP server allows your agent to bridge the gap between code and business logic, for example: Ensuring metric consistency: No more guessing whether a field is named total_revenue or revenue_total. Use get_explores and get_dimensions to ask your agent, "What is the correct measure for Net Retention?" and receive the precise field reference from the semantic model. Instantly validating logic: Don't wait to deploy a dashboard to test a theory. Use run_query to execute ad-hoc tests against the Looker model directly in your IDE, so that your application logic matches the live data. Auditing reports: Use run_look to pull results from existing saved reports, allowing you to verify that your application's output aligns with official business reporting. Build with data in Antigravity By integrating Google’s Data Cloud MCP servers into Antigravity, it’s easier than ever to use AI to discover insights and develop new applications. Now, with access to a wide variety of data sources that run your business, get ready to take the leap from simply talking to your code, to creating new experiences for your users. To get started, check out the following resources: Documentation: Connecting to AlloyDB using MCP GitHub: MCP Toolbox for Databases

When extreme weather or unexpected natural disaster strikes, time is the single most critical resource. For public sector agencies tasked with emergency management, the challenge isn't just about crafting a swift response, it’s about communicating that response to citizens effectively. At our recent Google Public Sector Summit, we demonstrated how Google Workspace with Gemini is helping government agencies turn complex, legally-precise official documents and text into actionable, personalized public safety tools almost instantly, thereby transforming the speed and efficacy of disaster response communication. Let’s dive deeper into how Google Workspace with Gemini can help transform government operations and boost the speed and effectiveness of critical public outreach during a natural disaster. The challenge: Turning authority into action Imagine you are a Communications Director at the Office of Emergency Management. In the aftermath of a severe weather event, the state government has just issued a critical Executive Order (EO), which serves as a foundational text, legally precise, and essential for internal agency coordination. However, its technical, authoritative language is not optimized for the public’s urgent questions such as: “Am I safe? Is my family safe? What should I do now?” Manually translating and contextualizing this information for the public, and finding official answers to critical questions - often hidden in the details - can create a dangerous information gap during a fast-moving natural disaster. Built on a foundation of trust Innovation requires security. Google Workspace with Gemini empowers agencies to adopt AI without compromising on safety or sovereignty, supported by: FedRAMP High authorization to meet the rigorous compliance standards of the public sector. Data residency & access controls including data regions, access transparency, and access approvals. Advanced defense mechanisms like context-aware access (CAA), data loss prevention (DLP), and client-side encryption (CSE). Operational resilience with Business Continuity editions to help keep your agency connected and operational during critical events. Google Workspace with Gemini: Your natural disaster response partner This is one area where Google Workspace with Gemini can help serve as your essential natural disaster partner, by empowering government leaders to move beyond manual translation and rapidly create dynamic, user-facing tools. For example, by using the Gemini app, the Communications Director at the Office of Emergency Management can simply upload the Executive Order PDF and prompt Gemini to 'create an interactive safety check tool based on these rules.' Gemini instantly parses the complex legal definitions—identifying specific counties, curfew times, and exemptions—and writes the necessary code to render a functional, interactive interface directly within the conversation window. What was once a static document becomes a clickable prototype in seconds, ready to be tested and deployed. Image: Gemini turns natural disaster declaration into an interactive map Three core capabilities driving transformation This process is driven by three core Google Workspace with Gemini capabilities. Unprecedented speed of transformation. The journey from a complex, static document to a working, interactive application is measured in minutes, not days or weeks. This acceleration completely changes the speed of development for mission-critical tools. In a disaster, the ability to deploy a targeted public safety resource instantly can be life-saving. Deep contextual understanding. Gemini's advanced AI goes beyond simple summarization. When provided with a full document and specific instructions, it can synthesize the data to perform complex tasks. For example, Gemini can analyze an executive order to identify embedded technical terms and locations, interpreting them as specific geographic areas that require attention. It extracts this pertinent information—while citing sources for grounding—and can transform raw text into a practical, location-aware tool for the public. A repeatable blueprint for any natural disaster. The entire process—from secure document upload to the creation of a working, live application—is repeatable. This means the model can be saved and leveraged for any future public safety resource, whether it’s a severe weather warning, a health advisory, or a general preparedness guide. This repeatable blueprint future-proofs an agency's ability to communicate quickly and effectively during any emergency. Serving the public with speed and clarity By leveraging Google Workspace with Gemini, public sector agencies can ensure that official emergency declarations immediately translate into clear, actionable details for the public. This shift from dense legal text to personalized guidance is paramount for strengthening public trust, improving citizen preparedness, and ultimately keeping communities safe. Are you ready to drive transformation within your own agency? Check out the highlights from our recent Google Public Sector Summit where leaders gathered to share how they are applying the latest Google AI and security technologies to solve complex challenges and advance their missions. Learn more about our Google Workspace Test Drive, and sign up for a no-cost 30-day pilot which provides your agency with full, hands-on access to the entire Google Workspace with Gemini, commitment-free, on your own terms.

We have exciting news for Google Cloud partners: Today we’re announcing our new partner program, Google Cloud Partner Network, which will formally roll out in the first quarter of 2026.  This new program marks a fundamental shift in how we measure success and reward value. Applicable to all partner types and sizes – ISVs, RSIs, GSIs, and more – the new program reinforces our strategic move toward recognizing partner contribution across the entire customer lifecycle.  Google Cloud Partner Network is being completely streamlined to focus on real-world results. This marks a strategic shift from measuring program work to valuing genuine customer outcomes. This includes rewarding successful co-sell sales efforts, high-quality service delivery, and shared innovation with ISVs. We are also integrating AI into the program’s core to make partner participation much easier, allowing more focus on customers instead of routine program administration. With its official kickoff in Q1, the new program will provide a six-month transition window for partners to adjust to the new framework. Today, we are sharing the first details of the Google Cloud Partner Network, which is centered on three pillars: simplicity, outcomes, and automation. Simplicity  We’re making the program simpler by moving away from tracking the work of traditional program requirements, such as business plans and customer stories, and towards recognising partner contributions – including pre-sales influence, co-innovation, and post-sales support.  Because the program is designed to put the customer first, we've narrowed requirements to focus on partner efforts that deliver real, measurable value. For example, the program will recognize investments in skills, real-world experience, and successful customer outcomes.  Outcomes  The new program will provide clear visibility into how partner impact is recognized and rewarded, focusing on customer outcomes. Critical highlights include:  New tiers: We’re evolving from a two-tier to a three-tier model: Select, Premier, and a new Diamond tier. Diamond is our highest distinction – it is intentionally selective, reserved for the few partners who consistently deliver exceptional customer outcomes. Each tier will now reflect our joint customer successes and will be determined based on exceptional customer outcomes across Google Cloud and Google Workspace. New baseline competencies: A new competency framework marks a fundamental shift that will replace today’s specializations, in order to reward partners for their deep technical and sales capabilities. The framework focuses on a partner’s proven ability to help customers, measuring two key dimensions: capacity (knowledge and skills development, validated by technical certifications and sales credentials) and capability (real-world success, measured by pre-sales and post-sales contributions to validated closed/won opportunities). This framework operates independently from tiering to allow partners to earn a competency without any dependency on a program tier. New advanced competency: The new global competencies introduce a second level of achievement, Advanced Competency, to signal a higher designation.  Automation Building on the proven success and transparency delivered through tools like the Earnings Hub and Statement of Work Analyzer, today’s Partner Network Hub will transform to deliver automation and transparency across the program.  The administrative responsibility for partners to participate in the program will be dramatically reduced through the use of AI and other tools. For example, a key change is the introduction of automated tracking across tiering and competency achievements. We will automatically apply every successful customer engagement toward a partner's progress in all eligible tiers and competencies. This radical simplification eliminates redundant reporting and ensures seamless, comprehensive recognition for the outcomes delivered. What’s next… The new program and portal will officially launch in Q1 2026, enabling partners to immediately log in, explore benefits and differentiation paths, and begin achieving new tiers and competencies. To ensure a smooth transition, we will host a series of webinars and listening sessions throughout early next year to educate partners on Google Cloud Partner Network. Stay tuned. We’ll have more to share soon!

When it comes to public health, having a clear picture of a community’s needs is vital. These insights help officials secure crucial funding, launch new initiatives, and ultimately improve people’s lives. That is the idea that inspired Dr. Phillip Levy, M.D., M.P.H., Professor of Emergency Medicine and Associate Vice President for Population Health and Translational Science at Wayne State University and his colleagues to develop Project PHOENIX: the Population Health OutcomEs aNd iNnformation eXchange. PHOENIX ingests information from electronic health records including demographic data, blood pressures and clinical diagnosis, and combines this with social and environmental factors from more than 70 anonymized data sources into an integrated virtual warehouse. Researchers, advocates,community leaders, and policy makers are able to use this data to better understand how different factors correlate to health outcomes and design targeted interventions. With such functionality, the PHOENIX team recognized the potential to transform the Community Health Needs Assessment (CHNA) process. Required by the federal government, public health departments, nonprofit hospitals, and Federally Qualified Health Centers in the United States must complete a CHNA every three years—a largely manual, time-consuming task that can take up to a year to complete. That’s where a collaboration between Wayne State University, Google Public Sector, and Syntasa came in. They teamed up to create CHNA 2.0, an innovative solution that drastically cuts down the time it takes to create these vital reports. By combining PHOENIX data with Vertex AI Platform, CHNA 2.0 can deliver a complete CHNA in a matter of weeks, giving health leaders valuable insights more quickly than ever. Extracting community sentiment from public data One of the most challenging parts of drafting a CHNA report involves conducting in-depth surveys to understand conditions in the community. This is often the most time-consuming part of the process, as it takes months to create, review, run, and analyze insightful surveys. By the time a CHNA report is complete, data from the surveys might be nearly a year out of date, which can prevent organizations from making a meaningful impact on their communities. CHNA 2.0 uses public health data from the PHOENIX warehouse along with insights from Syntasa Sentiment Analytics, which combines information from surveys with real-time data from Google Search and social media posts. Syntasa Sentiment Analytics provides insights regarding the questions people are asking and what issues they’re posting about to uncover health-related problems affecting a given community, such as growing concerns about asthma or frustrations with long waits at clinics. The architecture for this solution was built on the Syntasa Data + AI Platform. Workloads run on Google Kubernetes Engine (GKE) for its scalability, allowing the platform to process incoming sentiment data quickly. The platform also uses Cloud SQL and Google Cloud Storage as part of its data foundation, with BigQuery doing the heavy lifting for sentiment analysis. BigQuery provides the performance, efficiency, and versatility needed to handle large datasets of search and social media information efficiently. Creating reports with the power of humans + AI After gathering the necessary information, CHNA 2.0 uses Vertex AI and Gemini to help analysts create the report in less time. CHNA reports are highly complex and lengthy – and require manually integrating multiple data elements. Syntasa solved this challenge by breaking down the report into smaller, more manageable tasks and bringing human oversight into the loop. Now the person in charge of handling the CHNA defines the report’s structure. Gemini extracts insights from tailored datasets and fills in the relevant details. By combining both human and AI intelligence, CHNA 2.0 delivers reports in a fraction of the time. Organizations can also use this method to deliver a living document that is constantly updated with fresh data. This means public health officials don’t have to wait years to understand their communities—they can access the latest insights at any time to make faster and more impactful decisions.The net result is a transformation of the CHNA process from static to dynamic, enabling real time, data driven decision making for the betterment of all. Supporting public health with technology The City of Dearborn, Michigan, became the first to use CHNA 2.0 to great success. The long-term vision is to bring this same capability to other cities and counties in Michigan and across the nation. This project with Wayne State University and Syntasa showcases how the right technology and a strategic partner can create a powerful, scalable solution to a long-standing public sector challenge. By partnering with Google Public Sector to leverage the most advanced AI and data tools, Wayne State not only automated a critical process, but also empowered public health officials to better serve their communities. From improving community health to modernizing infrastructure, discover how technology is transforming the public sector. Sign up for our newsletter to stay informed on the latest trends and solutions.

Your security program is robust. Your audits are clean. But are you ready for a real-world attack? A tenacious human adversary can create a critical blind spot for security leaders: A program can be compliant, but not resilient. Bridging this gap requires more than just going through the red-teaming motions. To help security teams forge better instincts when responding to actual cyber-crisis events, we developed ThreatSpace, a cyber proving grounds and realistic corporate network that includes all the digital noise of real employee activities. From gaps to battle: The ThreatSpace cyber range The ThreatSpace environment is architecturally stateless and disposable to allow the deployment of real-world malware. It emulates the tactics, techniques, and procedures (TTPs) of real-world adversaries, informed by the latest, unparalleled threat intelligence from Google Threat Intelligence Group and Mandiant. By design, it never puts your actual business assets at risk. Watch Mandiant Red Teamers in action. See how our experts push the limits of red teaming. Because to stop a real attacker, you must think like one. Recently, stakeholders from the U.S. Embassy, the FBI, and Cote d'Ivoire cybersecurity agencies used ThreatSpace to conduct advanced defense training. Funded by the Bureau of International Narcotics and Law Enforcement Affairs (INL), this workshop brought together public and private sector partners to strengthen regional digital security. “Cybersecurity is a team sport, and our goal is to make Cote d'Ivoire a safer place for Ivorians and Americans to do business. This five-day workshop, funded by INL, brought together world-class instructors from Mandiant with local agencies and private sector partners to build the collaborative muscle we need to defend against modern threats," said Colin McGuire, FBI law enforcement attaché, Dakar in Cabo Verde and Gulf of Guinea. More than just helping to train individuals, we helped make the global digital ecosystem safer by uniting diverse groups of defenders facing shared threats. By practicing collaboration during a crisis, and operating as a unit, we can help empower defenders to fight and win against adversaries. ThreatSpace provides a safe place for your team to miss an indicator of compromise, exercise processes, and stress test collaboration and build the muscle memory and confidence needed to execute flawlessly when real adversaries come knocking. This is where an Offensive Security red team assessment comes in. Catch me if you can: The Mandiant red team reality check The Mandiant red team doesn’t follow a script. Our work on the frontlines of incident response lets us see precisely how determined adversaries operate, including their persistent, creative approaches to exploiting the complex seams between your technology, your processes, and your people. aside_block <ListValue: [StructValue([('title', 'Our Office of the CISO insights, direct to you'), ('body', <wagtail.rich_text.RichText object at 0x7fa3eb06e580>), ('btn_text', 'Subscribe today'), ('href', 'https://go.chronicle.security/cloudciso-newsletter-signup?utm_source=cgc-blog&utm_medium=blog&utm_campaign=FY23-Cloud-CISO-Perspectives-newsletter-blog-embed-CTA&utm_content=-&utm_term=-'), ('image', <GAEImage: Cloud CISO Perspectives new header July 2024 small>)])]> These observations enable our offensive security experts to mimic and emulate genuine threat actor behavior to achieve specific business objectives. Here are three scenarios developed by our red team to help stress-test and enhance our customers’ defenses: The "Impossible" Blackout. One organization believed their grid controls were isolated and secure. When our team demonstrated that a nationwide blackout was technically possible through their current architecture, the conversation shifted from compliance to survival. This finding empowered them to implement stricter controls immediately, preventing a theoretical catastrophe from becoming a reality. The Runaway Train. In another engagement, we gained remote system control of a locomotive train. The client didn't just get a technical report; they learned exactly how physical access vectors could bypass digital security. This exposure allowed them to harden their operational technology against vectors they had previously considered secure. The Generous Chatbot. Innovation brings new risks. In a recent test of a financial services chatbot, our team used simple prompts to bypass safety filters, ultimately convincing the AI to approve a 200-month loan at 0% APR. This finding prompted the client to immediately implement critical guardrails and grounding sources, ensuring they could innovate safely without exposing their business to manipulation. From reactive to resilient Building true cyber resilience requires a continuous feedback loop. It starts with analyzing your current state and enhancing your capability roadmap to align with operational priorities. Then you validate them through incident response learnings and offensive security insights and feed those back into the loop for the next iteration. By combining these disciplines, and grounding them with threat intelligence, you can move your organization from a reactive posture to a state of proactive resilience. You find and expose your weaknesses today, so you can build the strength required to secure your future. To battle-test your defenses, contact Mandiant to learn how our Offensive Security and ThreatSpace cyber range services can help you strengthen your defenses and build your resilience.

Today, we’re expanding the Gemini 3 model family with Gemini 3 Flash, which offers frontier intelligence built for speed at a fraction of the cost.  Gemini 3 Flash builds on the model series that developers and enterprises already love, optimized for high-frequency workflows that demand speed, without sacrificing quality. It allows enterprises to process near real-time information, automate complex workflows, and build responsive agentic applications. Gemini 3 Flash is built to be highly efficient, pushing the boundaries of quality at better price performance and faster speed. With a near real-time response from the model, businesses can now provide more engaging experiences for their end users at production scale, without sacrificing on quality.  Optimized for speed and scale Gemini 3 Flash strikes an ideal balance between reasoning and speed, for agentic coding, production-ready systems, and responsive interactive applications. It is available now in Gemini Enterprise, Vertex AI, and Gemini CLI, so businesses and developers can access:  Advanced multimodal processing: Gemini 3 Flash enables enterprises to build applications capable of complex video analysis, data extraction, and visual Q&As in near real-time. Whether streamlining back-office operations by extracting structured data from thousands of documents, or analyzing video archives to identify trends, Gemini 3 Flash delivers these insights with the speed required for modern data pipelines. Cost-efficient and high-performance execution for code and agents: Gemini 3 Flash delivers exceptional performance on coding and agentic tasks combined with a lower price point, allowing teams to deploy sophisticated reasoning across high-volume processes without hitting barriers. Low latency for near-real-time experiences: Gemini 3 Flash eliminates the lag typically associated with large models when it comes to intelligence. Its low latency powers responsive applications, from live customer support agents to in-game assistants. These applications can now offer more natural interactions for both quick answers and deep reasoning. Gemini 3 Flash clearly demonstrates that speed and scale do not have to come at the cost of intelligence. Real-world value across industries  With the launch of Gemini 3 Pro last month, we introduced frontier performance across complex reasoning, multimodal and vision understanding, as well as agentic and vibe-coding tasks. Gemini 3 Flash retains this foundation, combining Gemini 3's Pro-grade reasoning with Flash-level latency, efficiency, and cost. We are already seeing a tremendous response from companies using Gemini 3 Flash. With inference speed and reasoning capabilities that are typically associated with larger models, Gemini 3 Flash is unlocking new and more efficient use cases for companies like Salesforce, Workday and Figma. Reasoning and multimodality  "Gemini 3 Flash shows a relative improvement of 15% in overall accuracy compared to Gemini 2.5 Flash, delivering breakthrough precision on our hardest extraction tasks like handwriting, long-form contracts, and complex financial data. This is a significant jump in performance, and we're excited to continue collaborating to bring this specialist-level reasoning to Box AI users.” - Yashodha Bhavnani, Head of AI, Box "At Bridgewater, we require models capable of reasoning over vast, unstructured multimodal datasets without sacrificing conceptual understanding. Gemini 3 Flash is the first to deliver Pro-class depth at the speed and scale our workflows demand. Its long-context performance on complex problems is exceptional." - Jasjeet Sekhon, Chief Scientist and Head of AI, AIA Labs, Bridgewater Associates “ClickUp leverages Gemini 3 Flash's advanced reasoning to help power our next generation of autonomous agents. Gemini is decomposing high-level user goals into granular tasks, and we are seeing massive quality improvements on critical path identification and long-horizon task sequencing.” - Justin Midyet, Director, Software Engineering, ClickUp "Gemini 3 Flash has achieved a meaningful step up in reasoning, improving over 7% on Harvey's BigLaw Bench from its predecessor, Gemini 2.5 Flash. These quality improvements, combined with Flash's low latency, are impactful for high-volume legal tasks such as extracting defined terms and cross-references from contracts." - Niko Grupen, Head of Applied Research, Harvey Agentic coding "Our engineers have found Gemini 3 Flash to work well together with Debug Mode in Cursor. Flash is fast and accurate at investigating issues and finding the root cause of bugs." - Lee Robinson, VP of Developer Experience, Cursor “Gemini 3 Flash is a major step above other models in its speed class when it comes to instruction following and intelligence. It's immediately become our go-to for latency-sensitive experiences in Devin, and we're excited to roll it out to more use cases.” - Walden Yan, Co-Founder, Cognition "The improvements in the latest Gemini 3 Flash model are impressive. Even without specific optimization, we saw an immediate 10% baseline improvement on agentic coding tasks, including complex user-driven queries." - Daniel Lewis, Distinguished Data Scientist, Geotab “In our JetBrains AI Chat and Junie agentic-coding evaluation, Gemini 3 Flash delivered quality close to Gemini 3 Pro, while offering significantly lower inference latency and cost. In a quota-constrained production setup, it consistently stays within per-customer credit budgets, allowing complex multi-step agents to remain fast, predictable, and scalable.” - Denis Shiryaev, Head of AI DevTools Ecosystem, JetBrains “For the first time, Gemini 3 Flash combines speed and affordability with enough capability to power the core loop of a coding agent. We were impressed by its tool usage performance, as well as its strong design and coding skills.” - Michele Catasta, President & Head of AI, Replit "Gemini 3 Flash remains the best fit for Warp’s Suggested Code Diffs, where low latency and cost efficiency are hard constraints. With this release, it resolves a broader set of common command-line errors while staying fast and economical. In our internal evaluations, we’ve seen an 8% lift in fix accuracy." - Zach Lloyd, Founder & CEO, Warp Agentic applications   “Gemini 3 Flash is a great option for teams who want to quickly test and iterate on product ideas in Figma Make. The model can rapidly and reliably create prototypes while maintaining attention to detail and responding to specific design direction.” - Loredana Crisan, Chief Design Officer, Figma "Presentations.ai is using Gemini 3 Flash to enhance our intelligent slide-generation agents, and we’re consistently impressed by the Pro-level quality at lightning-fast speeds. With previous Flash-sized models there were many things we simply couldn’t attempt because of the speed vs. quality tradeoff. With Gemini 3 Flash, we’re finally able to explore those workflows.” - Saravanan Govindaraj, Co-Founder & Head of Product Development, Presentations.ai "Integrating Gemini 3 Flash into Agentforce is another step forward in our commitment to bring the best AI to our customers and deploy intelligent agents faster than ever. By pairing Google’s latest model capabilities with the power of Agentforce, we’re unlocking high-quality reasoning, stronger responses, and rapid iteration all inside the tools our customers already use." - John Kucera, SVP of Product Management, Salesforce AI "Gemini 3 Flash gives us a powerful new frontier model to fuel Workday's AI-first strategy. From delivering sharper inference in our customer-facing applications to unlocking greater efficiency in our own operations and development, it provides the performance boost to continue to innovate rapidly." - Dean Arnold, VP of AI Platform, Workday “Gemini 3 Flash model’s superb speed and quality allow our users to keep generating content without interruptions. With its improved Korean abilities and adherence to prompts, Gemini 3 Flash can be used for a variety of use cases including agentic workflow and story generation. As the largest consumer AI company in Korea, we’d love to keep using Gemini 3 models and be part of its continuous improvement cycles.” - DJ Lee, Chief Product Officer, WRTN Technologies Inc. Get started with Gemini 3 Flash Today, you can safely put Gemini 3 Flash to work. Business teams can access Gemini 3 Flash in preview on Gemini Enterprise, our advanced agentic platform for teams to discover, create, share, and run AI agents all in one secure platform.  Developers can start building with Gemini 3 Flash in preview on Vertex AI today. Gemini 3 Flash is also available in Google Antigravity, Gemini CLI, AI Studio, and more.

For most organizations, the question is no longer if they will use AI, but how to scale it from a promising prototype into a production-grade service that drives business outcomes. In this age of inference, competitive advantage is defined by your ability to serve useful information to users around the world at the lowest possible cost. As you move from demos to production deployments at scale, you need to simplify infrastructure operations with integrated systems that provide the latest AI software and accelerator hardware platforms, while keeping costs and architectural complexity low.  Yesterday, Forrester released The Forrester Wave™: AI Infrastructure Solutions, Q4 2025 report, evaluating 13 vendors, and we believe their findings validate our commitment to solving these core challenges. Google received the highest score of all vendors in the Current Offering category and received the highest possible score in 16 out of 19 evaluation criteria, including, but not limited to: Vision, Architecture, Training, Inferencing, Efficiency, and Security. Access the full report: The Forrester Wave™: AI Infrastructure Solutions, Q4 2025 Accelerating time-to-value with an integrated system Enterprises don’t run AI in a vacuum. They need to integrate it with a diverse range of applications and databases while adhering to stringent security protocols. Forrester recognized Google Cloud’s strategy of co-design by giving us the highest possible score in the Efficiency and Scalability criteria: “Google pursues a strategy of silicon-infrastructure co-design. It develops TPUs to improve inference efficiency and NVIDIA GPUs for access to broader ecosystem compatibility. Google designs TPUs to integrate tightly with its networking fabric, giving customers high bandwidth and low latency for inference at scale.” For over two decades, we have operated some of the world's largest services, from Google Search and YouTube to Maps, where their unprecedented scale required us to solve problems that no one else had. We couldn't simply buy the platform and infrastructure we needed; we had to invent it. This led to a decade-long journey of deep, system-level co-design, building everything from our custom network fabric and specialized accelerators to frontier models, all under one roof.  The result was an integrated supercomputing system, AI Hypercomputer, which has paid significant dividends for our customers. It supports a wide range of AI-optimized hardware, allowing you to optimize for granular, workload-level objectives — whether that's higher throughput, lower latency, faster time-to-results, or lower TCO. That means you can use our custom Tensor Processing Units (TPUs), the latest NVIDIA GPUs, or both, backed by a system that tightly integrates accelerators with networking and storage for exceptional performance and efficiency. It’s also why today, leading generative AI companies such as Anthropic, Lightricks, and LG AI Research trust Google Cloud to power their most demanding AI workloads.1 This system-level integration lays the foundation for speed, but operational complexity could still slow you down. To accelerate your time-to-market, we provide multiple ways to deploy and manage AI infrastructure, abstracting away the heavy lifting regardless of your preferred workflow. Google Kubernetes Engine (GKE) Autopilot automates management for containerized applications, helping customers like LiveX.AI reduce operational costs by 66%. Similarly, Cluster Director simplifies deployment for Slurm-based environments, enabling customers like LG AI Research to slash setup time from 10 days to under one day.  Managing AI cost and complexity Forrester gave Google Cloud the highest scores possible in the Pricing Flexibility and Transparency criterion. The price of compute is only one part of the AI infrastructure cost equation. A complete view should also account for development costs, downtime and inefficient resource utilization. We offer optionality at every layer of the stack to provide the flexibility businesses demand. Flexible consumption: Dynamic Workload Scheduler allows you to secure compute at up to 50% savings, by ensuring you only pay for the capacity you need, when you need it. Load balancing: GKE Inference Gateway improves throughput by using AI-aware routing to balance requests across models, preventing bottlenecks and ensuring servers aren't sitting idle. Eliminating data bottlenecks: Anywhere Cache co-locates data with compute, reducing read latency by up to 96% and eliminating the "integration tax" of moving data. By using Anywhere Cache together with our unified data platform BigQuery, you can avoid latency and egress fees while keeping your accelerators fed with data.  Mitigating strategic risk through flexibility and choice We are also committed to enabling customer choice across accelerators, frameworks and multicloud environments. This isn’t new for us. Our deep experience with Kubernetes, which we developed then open-sourced, taught us that open ecosystems are the fastest path to innovation and provide our customers with the most flexibility. We are bringing that same ethos to the AI era by actively contributing to the tools you already use. Open source frameworks and hardware portability: We continue to support open frameworks such as PyTorch, JAX, and Keras. We’ve also directly addressed concerns about workload portability on custom silicon by investing in TPU support for vLLM, allowing developers to easily switch between TPUs and GPUs (or use both) with only minimal configuration changes. Hybrid and multicloud flexibility: Our commitment to choice extends to where you run your applications. Google Distributed Cloud brings our services to on-premises, edge and cloud locations, while Cross-Cloud Network securely connects applications and users with high-speed connectivity between your environments and other clouds. This powerful combination means you're no longer locked into a specific environment; you can easily migrate workloads and apply uniform management practices, streamlining operations, and mitigating the risk of lock-in. Systems you can rely on When your entire business model depends on the availability of AI services, infrastructure uptime is critical. Google Cloud's global infrastructure is engineered for enterprise-grade reliability, an approach rooted in our history as the birthplace of Site Reliability Engineering (SRE). We operate one of the world's largest private software-defined networks, handling approximately 25% of global internet egress traffic. Unlike providers that rely on the public internet, we keep your traffic on Google’s own fiber to improve speed, reliability, and latency. This global backbone is powered by our Jupiter data center fabric, which scales to 13 Petabits/sec of bandwidth, delivering 50x greater reliability than previous generations — to say nothing of other providers. Finally, to improve cluster-level fault tolerance, we employ capabilities like elastic training and multi-tier checkpointing, which allow jobs to continue uninterrupted, by dynamically resizing the cluster around failed nodes while minimizing the time to recovery. Building on a secure foundation Our approach is to secure AI from the ground up. In fact, Google Cloud maintains a leading track record for cloud security. Independent analysis from cloudvulndb.org (2024-2025) shows that our platform has up to 70% fewer critical and high vulnerabilities compared to the other two leading cloud providers. We were also the first in the industry to publish an AI/ML Privacy Commitment, which guarantees that we do not use your data to train our models. With those safeguards in place, security is integrated into the foundation of Google Cloud, based on the zero-trust principles that protect Google’s own services: A hardware root of trust: Our custom Titan chips, as part of our Titanium architecture, create a verifiable hardware root of trust. We recently extended this with Titanium Intelligence Enclaves for Private AI Compute, allowing you to process sensitive data in a hardened, isolated, and encrypted environment. Built-in AI security: Security Command Center (SCC) natively integrates with our infrastructure, providing AI Protection by automatically discovering assets, preventing security issues, detecting active threats with frontline Google Threat Intelligence, and discovering known and unknown risks before attackers can exploit them.   Sovereign solutions: We enable you to meet stringent data residency, operational control, and software sovereignty requirements through solutions like Data Boundary. This is complemented by flexible options like partner-operated sovereign controls and Google Distributed Cloud for air-gapped needs. Platform controls for AI and agent governance: Vertex AI provides the essential governance layer for the enterprise builder to deploy models and agents at scale. This trust is anchored in Google Cloud’s secure-by-default infrastructure, utilizing platform controls like VPC Service Controls (VPC-SC) and Customer-Managed Encryption Keys (CMEK) to sandbox environments and protect sensitive data, and Agent Identity for granular IAM permissions. At the platform level, Vertex AI and Agent Builder integrate Model Armor to provide runtime protection against emergent agentic threats, such as prompt injection and data exfiltration.  Delivering continuous AI innovation We are honored to be recognized as a Leader in The Forrester Wave™ report, which we believe validates decades of R&D and our approach to building ultra-scale AI infrastructure. Look to us to continue on this path of system-level innovation as we help you convert the promise of AI into a reality. Access the full report: The Forrester Wave™: AI Infrastructure Solutions, Q4 2025 1. IDC Business Value Snapshot, Sponsored by Google Cloud, The Business Value of Google Cloud AI Hypercomputer, US53855425, October 2025

The complexity of the infrastructure behind AI training and high performance computing (HPC) workloads can really slow teams down. At Google Cloud, where we work with some of the world’s largest AI research teams, we see it everywhere we go: researchers hampered by complex configuration files, platform teams struggling to manage GPUs with home-grown scripts, and operational leads battling the constant, unpredictable hardware failures that derail multi-week training runs. Access to raw compute isn't enough. To operate at the cutting edge, you need reliability that survives hardware failures, orchestration that respects topology, and a lifecycle management strategy that adapts to evolving needs. Today, we are delivering on those requirements with the General Availability (GA) of Cluster Director and the Preview of Cluster Director support for Slurm on Google Kubernetes Engine (GKE).  Cluster Director (GA) is a managed infrastructure service designed to meet the rigorous demands of modern supercomputing. It replaces fragile DIY tooling with a robust topology-aware control plane that handles the entire lifecycle of Slurm clusters, from the first deployment to the thousandth training run.  We are expanding Cluster Director to support Slurm on GKE (Preview), designed to give you the best of both worlds: the familiar precision of high-performance scheduling and the automated scale of Kubernetes. It achieves this by treating GKE node pools as a direct compute resource for your Slurm cluster, allowing you to scale your workloads with Kubernetes' power without changing your existing Slurm workflows. Cluster Director, now GA Cluster Director offers advanced capabilities at each phase of the cluster lifecycle, spanning preparation (Day 0), where infrastructure design and capacity are determined; deployment (Day 1), where the cluster is automatically deployed and configured; and monitoring (Day 2), where performance, health, and optimization are continuously tracked. This holistic approach ensures that you get the benefits of fully configurable infrastructure while automating lower-level operations so your compute resources are always optimized, reliable, and available.  So, what does all this cost? That’s the best part. There's no extra charge to use Cluster Director. You only pay for the underlying Google Cloud resources — your compute, storage, and networking. How Cluster Director supports each phase of deployment Day 0: Preparation  Standing up a cluster typically involves weeks of planning, wrangling Terraform, and debugging the network. Cluster Director changes the ‘Day 0’ experience entirely, with tools for designing infrastructure topology that’s optimized for your workload requirements. To streamline your Day 0 setup, Cluster Director provides: Reference architectures: We’ve codified Google’s internal best practices into reusable cluster templates, enabling you to spin up standardized, validated clusters in minutes. This helps ensure that every team in your organization is using the same security standards for their deployments and deploying on infrastructure that is configured correctly by default — right down to the network topology and storage mounting.  Guided configuration: We know that having too many options can lead to configuration paralysis. The Cluster Director control plane guides you through a streamlined setup flow. You select your resources, and our system handles the complex backend mapping, ensuring that storage tiers, network fabrics, and compute shapes are compatible and optimized before you deploy. Broad hardware support: Cluster Director offers full support for large-scale AI systems, including Google Cloud’s A4X and A4X Max VMs powered by NVIDIA GB200 and GB300 GPUs, and versatile CPUs such as N2 VMs for cost-effective login nodes and debugging partitions. Flexible consumption options: Cluster Director integrates with your preferred procurement strategy, with support for Reservations for guaranteed capacity during critical training runs, Dynamic Workload Scheduler Flex-start for dynamic scaling, or Spot VMs for opportunistic low-cost runs. "Google Cloud's Cluster Director is optimized for managing large-scale AI and HPC environments. It complements the power and performance of NVIDIA's accelerated computing platform. Together, we're providing customers with a simplified, powerful, and scalable solution to tackle the next generation of computing challenges." - Dave Salvator, Director of Accelerated Computing Products, NVIDIA Day 1: Deployment  Deploying hardware is one thing, but maximizing performance is another thing entirely. Day 1 is the execution phase, where your configuration transforms into a fully operational cluster. The good news is that Cluster Director doesn't just provision VMs, it validates that your software and hardware components are healthy, properly networked, and ready to accept the first workload. To ensure a high-performance deployment, Cluster Director automates: Getting a clean "bill of health": Before your job ever touches a GPU, Cluster Director runs a rigorous suite of health checks, including DCGMI diagnostics and NCCL performance validation, to verify the integrity of the network, storage, and accelerators. Keeping accelerators fed with data: Storage throughput is often the silent killer of training efficiency. That’s why Cluster Director fully supports Google Cloud Managed Lustre with selectable performance tiers, allowing you to attach high-throughput parallel storage directly to your compute nodes, so your GPUs are never starved for data. Maximizing Interconnect Performance: To achieve peak scaling, Cluster Director implements topology-aware scheduling and compact placement policies. By utilizing dense reservations on Google’s non-blocking fabric, the system ensures that your distributed workloads are placed on the shortest physical path possible, minimizing tail latency and maximizing collective communication (NCCL) speeds from the get-go. “Cluster Director is an amazing product, which has enabled me to spin up a ready to use Nvidia GPU cluster with Slurm, including all networking, routing, and high performance network file-system for large-scale distributed model training within less than an hour. The cluster was immediately ready to run our containerizedAI training workloads with excellent throughput with only minimal customization effort." - Dr. Florian Eyben, Head of AI Foundation Models & Speech Technology, Agile Robots SE, Munich, Germany Day 2: Monitoring The reality of AI and HPC infrastructure is that hardware fails and requirements change. A rigid cluster is an inefficient cluster. As you move into the ongoing “Day 2” operational phase, you need to maintain cluster health, maximize utilization and performance. Cluster Director provides a control plane equipped for the complexities of long-term operations. Today we are introducing new active cluster management capabilities to handle the messy reality of Day 2 operations. New active cluster management capabilities include: Topology-level visibility: You can’t orchestrate what you can’t see. Cluster Director’s observability graphs and topology grids let you visualize your entire fleet, spot thermal throttles or interconnect issues, and optimize job placement based on physical proximity. One-click remediation: When a node degrades, you shouldn't have to SSH in to debug it. Cluster Director allows you to replace faulty nodes with a single click directly from the Google Cloud console. The system handles the draining, teardown, and replacement, returning your cluster to full capacity in minutes. Adaptive infrastructure: When your research needs change, so should your cluster. You can now modify active clusters, with activities such as adding or removing storage filesystems, on the fly, without tearing down the cluster or interrupting ongoing work. Cluster Director support for Slurm on GKE, now in preview Innovation thrives in the open. Google, the creator of Kubernetes, and SchedMD, the developers behind Slurm, have long championed the open-source technologies that power the world's most advanced computing. For years, NVIDIA and SchedMD have worked in lockstep to optimize GPU scheduling, introducing foundational features like the Generic Resource (GRES) framework and Multi-Instance GPU (MIG) support that are essential for modern AI. By acquiring SchedMD, NVIDIA is doubling down on its commitment to Slurm as a vendor-neutral standard, ensuring that the software powering the world's fastest supercomputers remains open, performant, and perfectly tuned for the future of accelerated computing. Building on this foundation of accelerated computing, Google is deepening its collaboration with SchedMD to answer a fundamental industry challenge: how to bridge the gap between cloud-native orchestration and high-performance scheduling. We are excited to announce the Preview of Cluster Director support for Slurm on GKE, utilizing SchedMD’s Slinky offering. This initiative brings together the two standards of the infrastructure world. By running a native Slurm cluster directly on top of GKE, we are amplifying the strengths of both communities: Researchers get the uncompromised Slurm interface and batch capabilities, such as sbatch and squeue, that have defined HPC for decades. Platform teams gain the operational velocity that GKE, with its auto-scaling, self-healing, and bin-packing, brings to the table. Slurm on GKE is strengthened by our long-standing partnership with SchedMD, which helps create a unified, open, and powerful foundation for the next generation of AI and HPC workloads. Request preview access now. Try Cluster Director today Ready to start using Cluster Director for your AI and HPC cluster automation?  Learn more about the end-to-end capabilities in documentation. Activate Cluster Director in the console.

December 15 - December 19 We introduced new self-service capabilities in Looker platform, enabling users to analyze local data alongside governed models, organize complex dashboards more effectively, and align the look and feel of their analytics with corporate branding. We outlined how you can connect your enterprise data to Google’s new Antigravity IDE. Leveraging the MCP Toolbox for Databases, you can securely connect your AI agents to services like AlloyDB for PostgreSQL, BigQuery, Spanner, Cloud SQL, Looker and others within Google’s Data Cloud, all within your development workflow. If you want trustworthy AI, what you need is a semantic layer that acts as the single source of truth for business metrics. We demonstrated how you can connect Looker to Gemini Enterprise in minutes, using the MCP Toolbox and Agent Development Kit (ADK). December 8 - December 12 We introduced Model Context Protocol (MCP) support for Google Services, starting with BigQuery, Maps, Google Compute Engine (GCE) and Google Kubernetes Engine (GKE). This announcement enables AI agents to point to a globally consistent and enterprise-ready endpoint for Google and Google Cloud services. In the next few months, we will roll out managed MCP support for more products, including AlloyDB, Cloud SQL, Spanner, Looker, Pub/Sub and Dataplex Universal Catalog. We recently launched the preview of data products in Dataplex Universal Catalog, Google Cloud’s unified, intelligent data to AI governance solution. A data product is a curated, ready-to-use package of data assets, documentation, and governance controls, all purposefully assembled to solve a specific business problem. Customers like Virgin Media O2 are using data products in Dataplex to deploy trusted data and scalable AI products. We recently commissioned  a Forrester Consulting Total Economic Impact™ (TEI) study. This report analyzes how building a data lakehouse with Google Cloud’s BigQuery and BigLake enabled organizations to get the flexibility of a data lake with open table formats like Apache Iceberg and the performance and governance of a high performance data warehouse, delivering the best of both on a single, open platform. Get the report to learn more about the ROI of building your data lakehouse with Google Cloud and how it can help you get your data AI-ready. December 1 - December 5 Last week, we launched AlloyDB AI’s forecasting capabilities. Traditionally, performing high-quality time-series forecasting (for sales, demand, inventory, etc.) required moving data to external platforms and engaging in long, complex model training and validation cycles. The new AI.FORECAST function solves this by bringing state-of-the-art predictive power directly into your operational database with a single SQL command. We announced the General Availability of Conversational Analytics in Looker, enabling all organizations to ask questions of of their data in natural language and get insightful answers, powered by Gemini. We followed up with the addition of Looker and Looker Conversational Analytics extensions in Gemini’s CLI. We held our Looker Innovation roadmap on December 4th. Hosted by Google Cloud product management and engineering leaders, you can view the recording on demand to learn the latest updates on Conversational Analytics and see how Looker is expanding self-service capabilities. We are making it easier for agent developers in Google’s Agent Development Kit (ADK) to answer questions. We are introducing BigQuery Agent Analytics, a new plugin for ADK that exports your agent's interaction data directly into BigQuery to capture, analyze, and visualize agent performance, user interaction, and cost. We are thrilled to announce the General Availability of AlloyDB AI’s columnar engine powered vector search. For critical applications like medical imaging, fraud detection, and legal research, 100% accuracy (perfect recall) is non-negotiable, meaning Approximate Nearest Neighbor (ANN) search is insufficient. However, running exact K-Nearest Neighbor (KNN) searches across large datasets often results in high latency, blocking real-time, mission-critical similarity searches. November 17 - November 21 This week, we launched a new, dedicated Cloud SQL Free Trial Instance program offering 30 days of risk-free access to Enterprise Plus features, including High Availability and Data Cache, on a powerful 8vCPU/64GB N2 instance. This is your chance to validate mission-critical performance for MySQL and PostgreSQL with zero operational commitment. (Docs) Our customers are leveraging the full power of Google Cloud's database portfolio to drive major business and technical breakthroughs. Companies like Pager Health and Dun & Bradstreet are using Cloud SQL and GKE to unify their infrastructure, reducing complexity while delivering world-class healthcare and risk solutions. For massive scale and emerging challenges, ID.me chose AlloyDB to handle 10TB+ workloads and power their generative AI projects with validated data to fight fraud for 145 million users. Finally, Palo Alto Networks built their globally distributed, high-availability security platform on Spanner and Spanner Graph, creating a robust data foundation for their critical, AI-driven workflows. November 3 - November 7  We have announced the next generation of Spanner-better-with-BigQuery capabilities delivering streaming insights, faster federated queries, cross-region data operations across Spanner and BigQuery data including Iceberg tables. Memory Agent for Cloud SQL for PostgreSQL is now generally available. Previously, memory-intensive queries could cause PostgreSQL restarts due to the Linux OOM killer. This led to downtime and no clear way for users to identify problematic queries. The new Memory Agent proactively detects and gracefully cancels high-memory connections, preventing restarts. With a recommender, it offers details and suggestions to alleviate memory pressure, providing a better user experience. We're excited to announce the General Availability of Customer-managed Active Directory integration with Cloud SQL for SQL Server. This allows Windows authentication for Cloud SQL for SQL Server instances using existing AD environments, eliminating the need for Google Managed AD and simplifying critical SQL Server workloads. October 24 - October 31 Dive into the newest Google Cloud Tech Bytes videos for Cloud SQL and Spanner! Get the practical details you need to set up and optimize our fully managed databases so you can simplify operations and accelerate development. October 20 - October 24 Database Migration Service now offers Object Level Observability, providing enhanced visibility and control over data migration. Previously limited to job-level oversight, these capabilities have been expanded to the individual table level, allowing for detailed insight into your data movement while heterogeneous database migration (e.g SQL Server to PostgreSQL). Cloud SQL's Enterprise Plus edition now supports the Axion-based C4A machine series in GA. This offers our customers significant performance benefits: nearly 50% better price-performance compared to current N2 machines and up to 2x greater transactional throughput than Amazon RDS Graviton 4-based offerings. Firestore with Enterprise Edition now offers Saved Queries.This new feature enables users to save and share queries for a specific database directly from the Firestore Studio page. At Oracle AI World ‘25, Database Center announced expanded support for Oracle Database@Google Cloud. This update allows customers to monitor Oracle Exadata and Autonomous databases, including their inventory and metrics, directly within the Database Center UI and Chat. Now, Google Cloud database services and Oracle inventory can be monitored side-by-side. Managed Kafka Connect is now generally available. Replicate on-prem clusters to Managed Service for Apache Kafka clusters, surface Kafka data in BigQuery, backup the data in Cloud Storage, or activate it in Pub/Sub. Unlock the real value of your Kafka data. Get started with Kafka Connect today. October 13 - October 17 Don't miss the Database Innovation Roadmap Webinar on October 30th, where we'll reveal the strategies and roadmap to supercharge agentic development and the next wave of AI innovation. This event kicks off our new Database Innovation Series, granting you access to 5+ deep-dive sessions shortly after the main event! October 6 - October 10 Cloud SQL now offers point-in-time recovery (PITR) for deleted instances, addressing compliance, accidental deletion, and disaster recovery needs. This feature requires customers to enable backup retention and PITR on their instances. Users can utilize the existing PITR clone API (with source-instance-deletion-time) and getLatestRecoveryTime API to manage deleted instances. The PITR window shortens based on log retention: up to 35 days for Enterprise Plus instances and 7 days for Enterprise instances. Introducing the Precheck API for Cloud SQL for PostgreSQL. This new feature improves Major Version Upgrades by proactively identifying potential issues, preventing unplanned downtime caused by instance incompatibilities (extensions, flags, data types). It addresses customer requests for a precheck utility to identify and remedy upgrade issues beforehand. AlloyDB now supports the tds_fdw extension, enabling direct access to SQL Server and Sybase databases. This feature streamlines database migrations and allows hybrid data analysis, complementing existing oracle_fdw support. September 29 - October 3 Cloud SQL announced support for Managed Connection Pool (in GA) across MySQL and PostgreSQL Managed Connection Pooling lets you scale your workloads by optimizing resource utilization for Cloud SQL instances using pooling. You can now also use IAM authentication to secure connections when using Managed Connection Pooling. To understand how it works, its key benefits, and how to configure Managed Connection Pooling for your workloads, dive into these guides: MySQL: https://discuss.google.dev/t/boost-your-cloud-sql-for-mysql-performance-through-managed-connection-pooling/269283 PostgreSQL: https://discuss.google.dev/t/optimizing-performance-and-scaling-with-managed-connection-pooling-for-cloud-sql-for-postgresql/270528?u=sagarsidhpura  September 22 - September 26 AlloyDB now supports PostgreSQL 17 in GA AlloyDB now offers general availability for PostgreSQL 17, bringing with it a range of new features and significant enhancements. Key improvements include: Improved query performance, particularly for materialized Common Table Expressions Incremental backup capabilities Enhanced logical replication features Improvements to the JSON data type handling Build AI Agents with Enterprise Databases (NEW! Training Course) This on-demand course teaches how to build AI agents that can leverage our enterprise databases using MCP Toolbox for Databases, as a secure middle layer. You will learn to securely connect AI agents to your existing databases like AlloyDB, Cloud SQL, and Spanner. You can define secure database interaction tools and implement intelligent querying capabilities, including semantic search with vector embeddings. Gemini CLI extensions for Data Cloud services and popular open source databases released In June, Google launched the open-source Gemini CLI. Now, developers can leverage open-source Gemini CLI extensions for Google Data Cloud services such as Cloud SQL, AlloyDB, and BigQuery. These extensions streamline data interactions and enhance application development directly from their local environment. For more details, check out the extensions documentation. You can also explore existing templates to begin creating and sharing your own extensions with the community. Cloud SQL for PostgreSQL now supports the pg_roaringbitmap extension Cloud SQL developers will now benefit from the ability to handle high-scale analytics, complex filtering, and large set operations directly within the managed PostgreSQL environment with unprecedented speed and efficiency. September 15 - September 19 Benchmark-Driven Kafka Optimization: Maximize Throughput and Cut Costs on Google Cloud Choosing the right compression strategy for Google Cloud Managed Service for Kafka is one of the most critical decisions impacting your performance and budget—and many are leaving massive savings on the table. Relying on default settings or guesswork can lead to unnecessarily high network and storage costs, increased latency, and severe throughput bottlenecks. This new, in-depth guide moves beyond theory to provide hard benchmark data, empowering you to make data-driven decisions.This comprehensive analysis systematically tests the most popular codecs (including GZIP, SNAPPY, and LZ4) against a "no compression" baseline. Read the full guide and get the sample benchmark code here. Explore and experiment with Spanner's advanced capabilities with ease. Say goodbye to friction and hello to innovation. Free 90-day trial Pre-loaded datasets for retail, banking, finance, and more Easy data import from MySQL, PostgreSQL dump files, and CSV Dozens of sample queries showcasing advanced features like full-text search, vector search, and graph capabilities C4A Axion processor support is now in GA for AlloyDB It was launched in Preview during Next'25. Customers waiting for GA to evaluate / onboard for production can now get better performance, price-performance and can run their development environment with 50% reduced entry price using one vCPU. Ready to get started? If you’re new to AlloyDB, You can sign-up via the AlloyDB free trial link. Parameterized Secured Views (now in Preview) in AlloyDB provides application data security and row access control using SQL views. September 8 - September 12 From query to cart: Inside Target’s search bar overhaul with AlloyDB AI  Target set out to modernize its digital search experience to better match guest expectations and support more intuitive discovery across millions of products. To meet that challenge, they rebuilt their platform with hybrid search powered by filtered vector queries and AlloyDB AI. Target achieved faster, smarter, more resilient search experience that’s already improved product discovery relevance by 20% and delivered measurable gains in performance and guest satisfaction. Powering smarter recommendations with Bigtable and BigQuery Schibsted Marketplaces, a leading online classifieds group in the Nordic region, cut infrastructure costs by 70% and accelerated data insights and model development by adopting Bigtable and BigQuery. This led to faster, more relevant recommendations and a better user experience. AlloyDB AI natural language support launched in Public Preview AlloyDB now simplifies the process for enterprises to develop highly accurate and secure Gen AI applications. These applications enable end-users to interact with their own data using natural language. The new natural language APIs integrate seamlessly into agentic architectures and are compatible with Gen AI orchestration frameworks like LangChain, making real-time operational data more accessible for end-user-facing chat experiences. Cloud SQL announced support for the Read Pools (in GA) across MySQL and PostgreSQL Cloud SQL's read pools offer a significant advantage over self-managed databases, particularly for read-heavy workloads. They simplify operations and enhance scalability by providing a single endpoint for up to 20 read pool nodes, automatically balancing traffic among them. Read pools can also be dynamically scaled up, down, out, or in to accommodate traffic surges. August 25 - August 29 Firestore with MongoDB compatibility is now generally available (GA)  Developers can now build cost-effective, scalable, and highly reliable apps on Firestore's serverless database using a familiar MongoDB-compatible API. With the general availability of Firestore with MongoDB compatibility, the 600,000 active developers within the Firestore community can now use existing MongoDB application code, drivers, and tools, as well as the open-source MongoDB ecosystem, with Firestore's serverless service. Firestore offers benefits like multi-region replication, virtually unlimited scalability, up to 99.999% SLA, single-digit millisecond read performance, integrated Google Cloud governance, and pay-as-you-go pricing.  Register now for an exciting  webinar on September 9th for a deep dive into Firestore with MongoDB compatibility and see live demos.  Database Migration Service (DMS) offers support for Private Service Connect (PSC) interfaces for homogenous migrations to Cloud SQL (PSCi support doc)  and AlloyDB (PSCi support doc). This capability is now generally available (GA). August 18 - August 22 Simplify Data Ingestion with the Revamped BigQuery "Add Data" Experience We're excited to announce the general availability of a completely redesigned "Add Data" experience in BigQuery, built to streamline how you bring data in for analysis. To enhance the user journey, we focused on simplifying the process of choosing from the many powerful ingestion methods BigQuery supports, from batch and streaming to CDC. Our goal was to create a more intuitive path for discovering data sources and provide clearer guidance on selecting the right tool for any given task. The new "Add Data" experience achieves this with a single, unified starting point within BigQuery Studio. It brings together all the ways to get data into BigQuery—including Data Transfer Service, Datastream, Dataflow, and partner solutions—into one intuitive interface. The experience guides you with clear categorization, solution recommendations, and in-context documentation to help you make informed choices. Now you can easily discover and configure the right data pipeline for your needs without leaving the BigQuery console. Get started by clicking the "+ Add data" button in the BigQuery Explorer pane today. Learn more in the official documentation. Cloud SQL now supports Private Service Connect (PSC) outbound connectivity With PSC outbound connectivity, customers can attach a PSC interface to their existing Cloud SQL PSC-enabled instances to allow their instances to make outbound connections to their network. This is required for homogeneous migrations using Database Migration Service. For more information, see PSC outbound connections. AI-Assisted Troubleshooting in Cloud SQL Enterprise Plus Cloud SQL for Enterprise Plus edition now offers enhanced AI-assisted troubleshooting, guiding you through resolving complex database performance issues such as slow queries and high load on your instances. This feature requires Gemini Cloud Assist and query insights, both available with the Enterprise Plus edition. August 11 - August 15 Code Your Way to $15,000: The BigQuery AI Hackathon Starts Now - go beyond traditional analytics and build groundbreaking solutions using BigQuery's cutting-edge AI capabilities. This is your opportunity to solve real-world business problems using BigQuery’s Generative AI, Vector Search, and Multimodal capabilities. You’ll get hands-on experience with BigQuery’s newest features that bring AI directly to your data. SQL users will find these capabilities feel like a natural extension of their existing workflow, while Python practitioners can use BigQuery DataFrames to work using a familiar, pandas-like API. The goal is simple: build powerful, scalable AI solutions right where your data lives. Sign-up today! AlloyDB now supports PG 17 (17.5 minor version) in Preview - AlloyDB customers can now access the latest improved version of Postgres, alongside existing versions like PG16, PG15, and PG14. Customers will also be able to upgrade to PG17 through MVU APIs. The community released PG17 in September 2024, introducing numerous new features and improvements. These include enhanced query performance (materialized Common Table Expressions, incremental backups and improved logical replication), a better developer experience (enhancements to the JSON support) and numerous other improvements.  Database Center now supports self-managed databases on GCE - Back in April, we announced the general availability of Database Center, your AI-powered unified fleet management solution for Google Cloud databases including Cloud SQL, AlloyDB, Spanner, Bigtable, Memorystore, and Firestore. However, many of our customers continue to leverage the flexibility of running their Postgres, MySQL and SQL server databases on Google Compute Engine (GCE) VMs. So we're thrilled to announce that Database Center now extends its monitoring capabilities to these self-managed databases. Please sign-up here to join this preview phase. Near Zero Downtime (nZDT) for Cloud SQL Enterprise Plus edition for SQL Server is now GA - With nZDT, maintenance and machine tier upgrades for Enterprise Plus SQL Server instances now experience sub-second downtime. This means: 99.99% SLA now includes maintenance downtime. Customers can say goodbye to lengthy planning cycles for maintenance. nZDT is now available across all three Cloud SQL engines - SQL Server, PostgreSQL and MySQL. Database Clone Feature in Firestore launched in Public Preview - Firestore database cloning allows Firestore users to create a copy of their database. All the Firestore Documents data, as well as index definitions and entries, are copied over to a new database in the same project & region with an appropriate user-chosen new database name. The user may choose to copy the state of the database from any snapshot time up to 7 days in the past. Build with Google Databases: 70+ Success Stories - This powerful resource highlights how over 70+ companies are using Google Cloud's fully managed database services to improve performance, scale globally, and optimize costs. It showcases real-world success stories across 10 industries, including retail, financial services, and technology. August 4 - August 8 Next Tokyo Data Cloud Announcements  - Google’s Data Cloud gives agents a complete, real-time understanding of your business, transforming it into a self-aware, reliable organization. We're delivering key innovations in three areas: 1) A new suite of data agents to act as expert partners, 2) An interconnected network for seamless agent collaboration, 3) A unified, AI-native foundation that unifies data and embeds AI-driven reasoning. AI-first Colab Enterprise experience in Vertex AI and BigQuery: This powerful platform streamlines complex data science workflows, allowing you to simply prompt an agent with a request like "train a model to predict income." The agent then autonomously generates and executes a complete plan—from data loading and cleaning to model training and evaluation Spanner Columnar Engine: Announcing the preview of the Spanner columnar engine, our latest innovation designed to turbocharge your data. By combining columnar storage and vectorized execution, we're making it possible to run lightning-fast analytical queries directly on your live operational data. Enhanced Backups for Cloud SQL: Introducing Enhanced Backups for Google Cloud SQL, now with logically air-gapped and immutable backup vaults. Built with Google Cloud Backup and DR Service, this is your ultimate defense against modern threats. July 28 - August 1 AlloyDB Omni now supports Kubernetes Operator 1.5.0 and PostgreSQL ver. 16.8.0/15.12.0: We have launched AlloyDB Omni Operator 1.5.0 and database versions 16.8.0/15.12.0. This major release delivers a significant step forward in enterprise readiness, including support for OpenShift operations, high availability/disaster recovery, and critical operational improvements like low-downtime upgrades and backups from standby. July 21 - July 25 Introducing partitioned index for BigQuery vector search: When creating a vector index on a partitioned BigQuery table, you now have the option to also partition your vector index. Partitioning your vector index significantly reduces query costs and improves search accuracy for vector searches that utilize pre-filtering on the partitioning column.By partitioning your vector index, BigQuery can apply partition pruning to both your table and your vector index when you use a filter on the partitioning column in your vector search. This means BigQuery only scans the relevant partitions, decreasing I/O costs. Additionally, pre-filtering on the partitioning column makes your vector searches less likely to miss relevant results. This feature is particularly beneficial if most of your vector searches target specific partitions using pre-filters. You can only partition TreeAH vector indexes, and the PARTITION BY clause used for the vector index must match the one used for the original table. Read more about the partitioned indexes in vector search. Datastream now supports BigLake Iceberg tables in BigQuery: Customers can now easily replicate data from different supported sources (MySQL, Postgres, SQLserver, Oracle,Salesforce and MongoDB ) of Datastream into BigLake Managed Tables for use cases spanning across open lakehouse, Enterprise grade storage for analytics, streaming and AI. Streaming to BigLake Iceberg tables lets you store data in a cost-effective way in the PARQUET format. By doing this, you can keep your data in a Cloud Storage bucket while using BigQuery for querying and analysis. Cloud SQL Write Endpoint for Advanced DR: Cloud SQL is excited to announce the GA of Write Endpoint to make Advanced Disaster Recovery (DR) seamless for customers (Documentation). This feature enhances application resilience during instance failovers and switchovers, ensuring customer applications remain connected to the primary instance without manual intervention.The write endpoint is now available in GA for MySQL and PostgreSQL instances of Enterprise Plus Edition. It already exists for SQL Server instances.  Vertical Scaling for Memorystore for Valkey and Memorystore for Redis Cluster: Using Vertical Scaling, Memorystore customers can now effortlessly scale their Memorystore nodes up or down ensuring optimal cluster sizing for varying workloads. Previously, node types were immutable post-deployment, hence customers only had the option for horizontal scaling (in and out) changing the number of shards in the cluster.  Database Migration Service (DMS) supports migrations from SQL Server to AlloyDB for PostgreSQL in GA: Customers can now use DMS to migrate their databases from SQL Server to AlloyDB for PostgreSQL . This migration offers seamless experience, which offers a comprehensive SQL Server modernization framework with: Automatic database schema and code conversion Gemini augmented database code conversion Gemini assisted PostgreSQL training and code improvements Low-downtime, CDC based data movement July 14 - July 18 Trust and security are central to Conversational Analytics. Designed to gain the benefits of Google’s most capable AI models, Conversational Analytics offers a powerful and insightful natural language experience that is secure and trustworthy, meaning you can realize the full potential of generative AI with confidence, while keeping your data under control. Learn more here. Turn questions into queries with the Conversational Analytics API. The Conversational Analytics API, now in preview, integrates multiple AI-powered tools to process user requests, including Natural Language to Query (NL2Query) and a Python code interpreter for generating responses, simplifying data science. Learn more here. Introducing BigQuery Soft Failover: Greater Control Over Disaster Recovery. BigQuery now offers "soft failover," giving administrators options over failover procedures. Unlike "hard failover" for unplanned outages, soft failover minimizes data loss for planned activities like disaster recovery drills or workload migrations. It initiates failover only after all data is replicated to the secondary region, guaranteeing data integrity. This feature is available via BigQuery UI, DDL, and CLI, providing enterprise-grade control for disaster recovery, confident simulations, and compliance without risking data. Learn more here.  July 7 - July 11 [Webinar] Join us for a session on "Build Smart Apps with Ease: Gen AI, Cloud SQL, and Observability for Faster Development." This webinar dives deep into mastering the essentials of building powerful Gen AI applications using Google Cloud technologies. Discover the complete Gen AI application development lifecycle, get a live demonstration of the new Application Design Center (ADC) for rapid app deployment, and explore its seamless integrations with frameworks like LangChain, LlamaIndex, and LangGraph. Plus, learn about the new MCP Toolbox for Databases to enhance the manageability and security of your GenAI agents, and understand critical operational considerations, including Cloud SQL Enterprise Plus features for performance, scalability, high availability, and disaster recovery. June 23 - June 27 Looker developers gain speed and accuracy with debut of Continuous Integration. Continuous Integration for Looker helps streamline code development workflows, boost the end-user experience, and gives developers the confidence to deploy changes faster. Learn more here. Code Interpreter brings advanced data science capabilities to Conversational Analytics. Code Interpreter helps answer complicated questions, tapping into Python to perform advanced analysis on your Looker data. Learn more here. June 16 - June 20 Standardize your business terminology with Dataplex business glossary. Want to standardize business terminologies and build a shared understanding across the enterprise? Dataplex business glossary is now GA within Dataplex Universal Catalog, providing a central, trusted vocabulary for your data assets, streamlining data discovery, and reducing ambiguity — leading to more accurate analysis, better governance, and faster insights. Learn more here. Looker Core on Google Cloud is now FedRAMP High authorized.  The need to protect highly sensitive government data is a top priority. Looker Core on Google Cloud enables users to explore and chat with their data via AI agents using natural language, and create dashboards and self-service reports. Learn more here. Fast Dev Mode Transition Speeds Looker Developers. A new Labs feature, Fast Dev Mode Transition, improves the performance of Development Mode on your Looker instance by loading LookML projects in read-only mode until a developer clicks the Create Developer Copy button for the project. Learn more here. Datastream now supports MongoDB as a Source (in Public Preview): You can now easily replicate data from MongoDB source into BigQuery and  Cloud Storage  for advanced analytics, reporting, and to power generative AI applications. Datastream offers MongoDB connectivity for both Replica Sets and Sharded Clusters. This includes support for self-managed MongoDB deployments as well as the fully managed AtlasDB service.  Private Service Connect (PSC) on existing Cloud SQL instances (GA): Cloud SQL now offers the ability to enable Private Service Connect (PSC) on existing instances that currently utilize Private Service Access (PSA). This new functionality, generally available for PostgreSQL, MySQL, and SQL Server engines, eliminates the previous requirement of creating new instances for PSC adoption. Customers can now transition their existing PSA instances to PSC without data migration.  Cloud SQL for SQL Server - E+ Recommender: The Enterprise Plus recommender helps customers identify SQL Server instances that would benefit from an upgrade to the Cloud SQL Enterprise Plus Edition. It offers insights into current performance metrics, and emphasizes how Enterprise Plus features (such as the data cache and memory-optimized machines) can boost performance. Additionally, the recommender includes a convenient button for direct navigation to the instance settings page, enabling users to perform the upgrade easily.  AlloyDB - PSC Service Automation: With this launch, AlloyDB significantly improves the connectivity configuration experience for Private Service Connect (PSC), by automatically creating PSC endpoints in the customer VPC and exposing the IP address of the endpoint directly through the AlloyDB API, enabling seamless PSC adoption at scale. June 9 - June 13 Introducing Pub/Sub Single Message Transforms (SMTs), to make it easy to perform simple data transformations such as validate, filter, enrich, and alter individual messages as they move in real time right within Pub/Sub. The first SMT is available now: JavaScript User-Defined Functions (UDFs), which allows you to perform simple, lightweight modifications to message attributes and/or the data directly within Pub/Sub via snippets of JavaScript code. Learn more in the launch blog.  Serverless Spark is now generally available directly within BigQuery. Formerly Dataproc Serverless, the fully managed Google Cloud Serverless for Apache Spark helps to reduce TCO, provides strong performance with the new Lightning Engine, integrates and leverages AI, and is enterprise-ready. And by bringing Apache Spark directly into BigQuery, you can now develop, run and deploy Spark code interactively in BigQuery Studio. Read all about it here.  Next-Gen data pipelines: Airflow 3 arrives on Google Cloud Composer: Google is the first hyperscaler to provide selected customers with access to Apache Airflow 3, integrated into our fully managed Cloud Composer 3 service. This is a significant step forward, allowing data teams to explore the next generation of workflow orchestration within a robust Google Cloud environment. Airflow 3 introduces powerful capabilities, including DAG versioning for enhanced auditability, scheduler-managed backfills for simpler historical data reprocessing, a modern React-based UI for more efficient operations, and many more features. June 2 - June 6 Enhancing BigQuery workload management: BigQuery workload management provides comprehensive control mechanisms to optimize workloads and resource allocation, preventing performance issues and resource contention, especially in high-volume environments. To make it even more useful, we announced several updates to BigQuery workload management around reservation fairness, predictability, flexibility and “securability,” new reservation labels, as well as autoscaler improvements. Get all the details here.  Bigtable Spark connector is now GA: The latest version of the Bigtable Spark connector opens up a world of possibilities for Bigtable and Apache Spark applications, not least of which is additional support for Bigtable and Apache Iceberg, the open table format for large analytical datasets. Learn how to use the Bigtable Spark connector to interact with data stored in Bigtable from Apache Spark, and delve into powerful use cases that leverage Apache Iceberg in this post. BigQuery gets transactional: Over the years, we’ve added several capabilities to BigQuery to bring near-real-time, transactional-style operations directly into your data warehouse, so you can handle common data management tasks more efficiently from within the BigQuery ecosystem. In this blog post, you can learn about three of them: efficient fine-grained DML mutations; change history support for updates and deletes; and real-time updates with DML over streaming data. Google Cloud databases integrate with MCP: We announced capabilities in MCP Toolbox for Databases (Toolbox) to make it easier to connect databases to AI assistants in your IDE. MCP Toolbox supports BigQuery, AlloyDB (including AlloyDB Omni), Cloud SQL for MySQL, Cloud SQL for PostgreSQL, Cloud SQL for SQL Server, Spanner, self-managed open-source databases including PostgreSQL, MySQL and SQLLite, as well as databases from other growing list of vendors including Neo4j, Dgraph, and more. Get all the details here.

At Google Cloud, we continue to make critical investments to Vertex AI Agent Builder, our comprehensive and open platform, enabling you to build faster, scale efficiently, and govern with enterprise-grade security. Today, with the integration of the Cloud API Registry, we’re excited to bring enhanced tool governance capabilities to Vertex AI Agent Builder. With this latest update, administrators can now manage available tools for developers across your organization directly in Vertex AI Agent Builder Console, and developers can leverage tools managed by the registry with a new ApiRegistry.  With this, organizations can anchor agents in the embedded security and operational controls that they already use, enabling deploying and managing agents as a digital workforce. Following last month's expansion of our Agent Builder platform, we are also introducing new capabilities across the entire agent lifecycle to help developers build faster using new ADK capabilities and visual tools, and scale with high performance through the expansion of Agent Engine services, including the general availability of support for sessions and memory. Read more below. 1. Govern your tools with confidence Building a useful agent requires the agent to have access to the necessary tools. However, developers today spend a significant amount of time building their tools for each agent, resulting in duplicate work. This approach also presents challenges for administrators who want to control what data and tools agents can access.  We are bringing enhanced tool governance with the integration of Cloud API Registry in the Vertex AI Agent Builder Console. This acts as a private registry that administrators can use to curate and govern a set of approved tools for developers to use across their organization, providing: Pre-built tools for Google services: We recently announced MCP support for Google services like BigQuery and Google Maps, which will be available for use in Vertex AI Agent Builder. Support for custom MCP servers: Unlock your entire API estate for the agentic age. Apigee now empowers you to transform your existing managed APIs into custom MCP servers, bridging your established digital assets with modern AI workflows. Additionally, by bringing these tools from multiple clouds into Apigee API hub, you help ensure your agent developers have instant and secure access to a curated catalog through the Cloud API Registry. Enhanced tool management: Administrators using the new experience in Vertex AI Agent Builder to view, govern, and manage tools can now ensure the right tools are available to developers in their organization. Simplified tool access: For developers, Agent Development Kit (ADK) introduces support of Cloud API Registry, introducing a new ApiRegistry object to easily leverage managed tools. The demo above showcases the new user journey for managing and governing tools directly within the Vertex AI Agent Builder Console 2. Build your AI agents faster Last month, we released Gemini 3 Pro, our most intelligent model, to every developer and enterprise team. It’s the best model in the world for multimodal understanding, and our most powerful agentic model yet. With full compatibility with ADK, you can now build, test, and deploy powerful AI agents with greater reliability and confidence. We are introducing new capabilities to help you move from concept to interactive product: Full ADK support of Gemini 3 Pro and Flash: ADK now fully supports Gemini 3 Pro and Flash, allowing you to build reliable, production-ready agents.  ADK for TypeScript: We are extending ADK support for TypeScript, ensuring you can leverage the latest capabilities in ADK directly in whatever language you choose.  State management in ADK: We've made significant improvements to our agentic state management within ADK, which is the system for an AI agent to maintain context and memory during and across conversations. New improvements include:  Recovery from failure: If a conversation crashes due to an error, ADK now restores the state natively, requiring no additional work from the developer. Continue with human-in-the-loop: You can now pause for human input anywhere, even inside complex workflows. ADK automatically remembers exactly where the agent stopped and resumes immediately after approval, so you don't have to write extra code to track progress. Rewind state and context: Developers can now rewind to any point in the conversation and invalidate all interactions after that point so the user can remove the “polluted” context rather than send a new message. This allows users to try different approaches to solving a problem without having to open new sessions. Interactions API integration: ADK and the Agent2Agent protocol (A2A) now support the new Interactions API, providing a consistent way to manage multimodal input/output (text, audio, visual) across your agents, simplifying integration with client applications. A2UI: Built on top of A2A protocol, A2UI is an early-stage UI toolkit to facilitate LLM-generated UIs for remote agents. This allows you to enable agents to pass shared UI widgets and components directly to user-facing applications without the security risks and overhead of iframes or sending executable code, allowing you to build rich user experiences securely. A2UI Landscape Architect Demo: Build with Vertex AI Agent Builder Above is a demo showcasing A2UI in action where the user uploads a photo, a remote agent uses Gemini to understand it, and dynamically generates a custom form using A2UI for the specific needs of the customer. You can start building today with adk-samples on GitHub or on Vertex AI Agent Garden, a growing repository of curated agent samples, solutions, and tools designed to accelerate your development and support one-click deployment of your agents built with ADK. Access our Agent Starter Pack, a template collection that provides a production-ready foundation for building, testing, and deploying AI agents.  3. Scale your AI agents effectively Once you’ve built your agent, the next challenge is going into a production environment. That’s why we continue to expand the managed services available in Agent Engine to provide the core capabilities needed to scale your agents. Manage context with confidence: We are moving Agent Engine sessions and memory bank to General Availability (GA). You can now use Agent Engine to manage both short-term and long-term memory for your production workloads. This allows your agents to maintain context across different interactions, which is critical for delivering helpful, personalized responses at scale. This product is powered by Google Cloud AI Research’s novel research method (accepted by ACL 2025), using a topic-based approach that sets a new standard for how agents learn and recall information. Expanded regional support for Agent Engine services: All Agent Engine services are now available in seven additional regions worldwide. To learn more, refer to the documentation. Pricing updates for Agent Engine: We lowered pricing for the Agent Engine runtime and will begin billing for additional Agent Engine services starting on January 28, 2026. You can review the Agent Engine pricing documentation for additional detail and hypothetical agent cost scenarios.   Product Resource SKU Prior pricing New pricing Price change date Runtime vCPU / hour 8A55-0B95-B7DC $0.0994 $0.0864 December 16, 2025 Memory / GB-hr 0B45-6103-6EC1 $0.0105 $0.0090 December 16, 2025 Code Execution vCPU / hour 448F-9419-C2EE Free $0.0864 January 28, 2026 Memory / GB-hr AC0F-52B0-CE44 Free $0.0090 January 28, 2026 Sessions Stored session events 0D5A-FCD2-CB63 Free $0.25/1,000 events January 28, 2026 Memory Bank Memories stored per month E954-622B-C859 Free $0.25/1,000 memories (LLMcostsbilled separately) January 28, 2026 Memories retrieved 6DEC-3026-DDFF Free $0.50/1,000 memories  January 28, 2026 The table above shows updated pricing for Agent Engine services and when the changes take place.  How customers are achieving more with Agent Builder "Burns & McDonnell uses Vertex AI Agent Builder to transform how organizational knowledge is applied across the enterprise. With Experience IQ, we are building an AI agent using ADK that turns decades of project data and employee experience into real-time, actionable intelligence. Vertex AI enables this innovation to scale responsibly by combining deterministic business rules with probabilistic reasoning, making AI a trusted operational capability — not just a productivity tool. This agent helps teams quickly identify the right experience, reduce manual effort in staffing and planning, and make higher-confidence decisions grounded in verified data. With Vertex AI, Burns & McDonnell isn’t just managing knowledge — we are activating experience to drive faster, more confident decisions." - Matt Olson, Chief Innovation Officer, Burns & McDonnell “Payhawk uses Vertex AI Agent Builder to transform agents into financial assistants that truly ‘know’ our customers. Leveraging Memory Bank, we moved from stateless interactions to long-term context retention, allowing agents to recall user constraints and historical patterns with continuity. For example, our Financial Controller Agent now remembers habits like expensing small meals and auto-submits them, reducing submission time by over 50%. Similarly, our Travel Agent proactively applies preferences like aisle seats. This significantly drops cognitive load, allowing agents to anticipate needs based on past behavior rather than just reacting to prompts.” - Diyan Bogdanov, Principal Applied AI Engineer, Payhawk "Gurunavi uses Vertex AI Agent Builder to power 'UMAME!', an AI restaurant discovery app that leverages Agent Engine's Memory Bank to overcome a significant challenge: achieving a deep understanding of user context. Unlike conventional prompt-based systems, our agent leverages memory bank to remember a user's past actions, preferences, and temporal patterns to proactively present the best options. This eliminates the need for manual searches, creating a seamless experience. We project this context-aware capability will improve user experience by 30% or more. We view this memory function as a non-negotiable feature for helping everyone forge new culinary experiences together with AI.” - Toshiaki Iwamoto, CTO, Gurunavi “SeaArt Entertainment uses Vertex AI Agent Builder to personalize the creative experience for digital artists. Before Memory Bank, our AI agents could not reliably remember users’ preferences. For example, when users worked on complex multimodal art projects, they had to repeatedly explain the same details — like their favorite character styles or model choices — across sessions. After integrating Memory Bank, our agents are now able to recall past conversations, actions, and user preferences. We especially like that the agent can seamlessly remember context across sessions, making interactions feel more natural and personal." - Aleksei Savin, Lead of Multimodal AI Platform, SeaArt Entertainment Get started Vertex AI Agent Builder provides the unified platform to manage the entire agent lifecycle, helping you close the gap from prototype to a production-ready agent. To explore these new features, visit the updated Agent Builder documentation and release notes. If you’re a startup and you’re interested in learning more about building and deploying agents, download the Startup Technical Guide: AI Agents. This guide provides the knowledge needed to go from an idea to prototype to scale, whether your goals are to automate tasks, enhance creativity, or launch entirely new user experiences for your startup.

Today’s AI capabilities provide a great opportunity to enable natural language (NL) interactions with your enterprise data through applications using text and voice. In fact, in the world of agentic applications, natural language is rapidly becoming the interaction standard. That means agents need to be able to issue natural language questions to a database and receive accurate answers in return. At Google Cloud, this drove us to build Natural-Language-to-SQL (NL2SQL) technology in the AlloyDB database that can receive a question as input and return a NL result, or the SQL query that will help you retrieve it. Currently in preview, the AlloyDB AI natural language API enables developers to build an agentic application that answers natural language questions on their database data by agents or end users in a secure, business-relevant, explainable manner, with accuracy approaching 100% — and we’re focused on bringing this capability to a broader set of Google Cloud databases.  When we first released the API in 2024, it already provided leading NL2SQL accuracy, albeit not close to 100%. But leading accuracy isn’t enough. In many industries, it’s not sufficient to translate text into SQL with accuracy of 80% or even 90%. Low-quality answers carry a real cost, often measurable in monetary terms: disappointed customers or poor business decisions. A real estate search application that fails to understand what the end user is asking for (their “intent”) risks becoming irrelevant. In retail product search, less relevant answers lead to lower conversions into sales. In other words, the accuracy of the text-to-SQL translation must almost always be extremely high. In this blog we help you understand the value of the AlloyDB AI natural language API and techniques for maximizing the accuracy of its answers. Getting to ~100% accurate and relevant results Achieving highly accurate text-to-SQL takes more than just prompting Gemini with a question. Rather, when developing your app, you need to provide AlloyDB AI with descriptive context, including descriptions of the database tables and columns; this context can be autogenerated. Then, when the AlloyDB AI natural language API receives a question, it can intelligently retrieve the relevant pieces of descriptive context, enabling Gemini to see how the question relates to the database data. Still, many of our customers asked us for explainable, certifiable and business-relevant answers that would enable them to reach even higher accuracy, approaching 100% (such as >95% or even higher than 99%), for their use cases.  The latest preview release of the AlloyDB AI natural language API provides capabilities for improving your answers in several ways: Business relevance: Answers should contain and properly rank information in order to improve business metrics, such as conversions or end-user engagement. Explainability: Results should include an explanation of intent that clarifies — in language that end users can understand — what the NL API understood the question to be. For example, when a real estate app interprets the question “Can you show me Del Mar homes for families?” as “Del Mar homes that are close to good schools”, it explains its interpretation to the end user. Verified results: The result should always be consistent with the intent, as it was explained to the user or agent. Accuracy: The result should correctly capture the intent of the question.  With this, the AlloyDB AI natural language API enables you to progressively improve accuracy for your use case, what’s sometimes referred to as “hill-climbing”. As you work your way towards 100% accuracy, AlloyDB AI’s intent explanations mitigate the effect of the occasional remaining inaccuracies, allowing the end user or agent to understand that the API answered a slightly different question than the one they intended to ask. aside_block <ListValue: [StructValue([('title', 'Get started with a 30-day AlloyDB free trial instance'), ('body', <wagtail.rich_text.RichText object at 0x7fa3e844ed00>), ('btn_text', ''), ('href', ''), ('image', None)])]> Hill-climbing to approximate 100% accuracy  Iteratively improving the accuracy of AlloyDB AI happens via a simple workflow. First, you start with the NL2SQL API that AlloyDB AI provides out of the box. It’s highly (although not perfectly) accurate thanks to its built-in agent that translates natural language questions into SQL queries,  as well as automatically generated descriptive context that is used by the included agent.  Next, you can quickly iterate to hill-climb to approximately 100% accuracy and business relevance by improving context. Crucially, in the AlloyDB AI natural language API, context comes in two forms:  Descriptive context, which includes table and column descriptions, and  Prescriptive context, which includes SQL templates and (condition) facets, allowing you to control how the NL request is translated to SQL.  Finally, a “value index” disambiguates terms (such as SKUs and employee names) that are private to your database, and thus that are not immediately clear to foundation models. The ability to hill-climb to approximate 100% accuracy flexibly and securely relies on two types of context and the value index in AlloyDB. Let’s take a deeper look at context and the value index. 1. Descriptive and prescriptive context  As mentioned above, the AlloyDB AI natural language API relies on descriptive and prescriptive context to improve the accuracy of the SQL code it generates. By improving descriptive context, mostly table and column descriptions, you increase the chances that the SQL queries employ the right tables and columns in the right roles. However, prescriptive context resolves a harder problem: accurately interpreting difficult questions that matter for a given use case. For example, an agentic real-estate application may need to answer a question such as “Can you show me homes near good schools in <provided city>?” Notice the challenges: What exactly is “near”? How do you define a “good” school? Assuming the database provides ratings, what is the cutoff for a good school rating? What is the optimal tradeoff (for ranking purposes and thus for business relevance of the top results) between distance from the school and ranking of the school when the solutions are presented as a list? To help, the AlloyDB natural language API lets you supply templates, which allow you to associate a type of question with a parameterized SQL query and a parameterized explanation. This enables the AlloyDB NL API to accurately interpret natural language questions that may be very nuanced; this makes templates a good option for frequently asked, nuanced questions. A second type of prescriptive context, facets, allows you to provide individual SQL conditions along with their natural language counterparts. Facets enable you to combine the accuracy of templates with the flexibility of searching over a gigantic number of conditions. For example, “near good schools” is just one of many conditions. Others may be price, “good for a young family”, “ocean view” or others. Some are combinations of these conditions, such as “homes near good schools with ocean views”. But you can’t have a template for each combination of conditions. In the past, to accommodate all these conditions, you could have tried to create a dashboard with a search field for every conceivable condition, but it would have become very unwieldy, very fast. Instead, when you use a natural language interface, you can use facets to cover any number of conditions, even in a single search field. This is where the strength of a natural language interface really shines!  The AlloyDB AI natural language API facilitates the creation of descriptive and prescriptive context. For example, rather than providing parameterized questions, parameterized intent explanations, and parameterized SQL, just add a template via the add_template API, in which you provide an example question (“Del Mar homes close to good schools”) and the correct corresponding SQL. AlloyDB AI automatically generalizes this question to handle any city and automatically prepares an intent explanation. 2. The value index The second key enabler of approximate 100% accuracy is the AlloyDB AI value index, which disambiguates terms that are private to your database and, thus, not known to the underlying foundation model. Private terms in natural language questions pose many problems. For starters, users misspell words, and, indeed, misspellings increase with a voice interface. Second, natural language questions don’t always spell out a private term’s entity type. For instance, a university administrator may ask “How did John Smith perform in 2025?” without specifying whether John Smith is faculty or a student; each case requires a different SQL query to answer the question. The value index clarifies what kind of entity “John Smith” is, and can be automatically created by AlloyDB AI for your application. Natural language search over structured, unstructured and multimodal data  When it comes to applications that provide search over structured data, the AlloyDB AI natural language API enables a clean and powerful search experience. Traditionally, applications present conditions as filters in the user interface that the end user can employ to narrow their search. In contrast, an NL-enabled application can provide a simple chat interface or even take voice commands that directly or indirectly pose any combination of search conditions, and still answer the question. Once search breaks free from the limitations of traditional apps, the possibilities for completely new user experiences really open up. The combination of the NL2SQL technology with AI search features also makes it good for querying combinations of structured, unstructured and multimodal data.The AlloyDB AI natural language API can generate SQL queries that include vector search, text search and other AI search features such as the AI.IF condition, which enables checking semantic conditions on text and multimodal data. For example, our real estate app may be asked about “Del Mar move-in ready houses”. This would result in a SQL query with an AI.IF function that checks whether the text in the description column of the real_estate.properties table is similar to “move-in ready”. Bringing the AlloyDB AI natural language API into your agentic application  Ready to integrate the AlloyDB AI natural language API into your agentic application? If you’re writing AI tools (functions) to retrieve data from AlloyDB, give MCP Toolbox for Databases a try. Or for no-code agentic programming, you can use Gemini Enterprise. For example, you can create a conversational agentic application that uses Gemini to answer questions from its knowledge of the web and the data it draws from your database — all without writing a single line of code! Either way, we look forward to seeing what you build.

AI and cloud technology are reshaping every corner of every industry around the world. Without our customers, there would be no Google Cloud, as they are the ones building the future on our platform. In this regular round-up, we dive into some of the exciting projects redefining businesses, shaping industries, and creating new categories.  For our latest edition, we look into how Waze made its network more reliable; NBA superstar Stephen Curry gets quizzed by Gemini; a financial market transformation at CME Group; a multi-agent business forecasting platform from AppOrchid; Mattel crunches customer feedback with AI; VMO2 uses decentralized contracts for reliable data; Mercado Libre’s strategic use of Spanner; and how Ericsson enhances data governance. Be sure to check back next year to see how more industry leaders and exciting startups are putting Google Cloud technologies to use. And if you haven’t already, please peruse our list of 1,001 real-world gen AI use cases from our customers. Waze keeps traffic flowing with Memorystore Who: Waze (a division of Google parent company Alphabet) is a community-driven, crowd-sourced navigation app with tens of millions of users who share real-time data to provide optimal driving routes, traffic updates, and alerts for hazards, police, and more. What they did: Waze depends on vast volumes of dynamic, real-time user session data to power its core navigation features, but scaling that data to support concurrent users worldwide required a new approach. Their team built a centralized Session Server backed by Memorystore for Redis Cluster, a fully managed service with 99.99% availability that supports partial updates and easily scales to Waze’s use case of over 1 million MGET commands per second with ~1ms latency. Why it matters: Moving from Memcached’s 99.9% SLA to Memorystore for Redis Cluster’s 99.99% means higher availability and resiliency from the service. And because Memorystore for Redis supports partial updates, Waze can change individual fields within a session object rather than rewriting the entire record. That reduces network traffic, speeds up write performance, and makes the system more efficient overall. Learn from us: “Real-time data drives the Waze app experience. Our turn-by-turn guidance, accident rerouting, and driver alerts depend on up-to-the-millisecond accuracy. But keeping that experience seamless for millions of concurrent sessions requires robust and battle hardened infrastructure that is built to manage a massive stream of user session data.” – Eden Levin, Waze BE infrastructure developer & Yuval Kamran Waze site reliability engineer What Stephen Curry learned from a custom Gemini agent Who: Stephen Curry is arguably of the greatest three-point shooter of all-time in the NBA — as well as Google’s performance advisor and an all-around stats-obsessive. What they did: For a special engagement with Curry, the Google Cloud team wanted to showcase the power of Gemini for creative thinking, analysis, and data mining. They took every regular season, play-in, and playoff game from Curry’s career (through the end of the 2024-2025 season) and input the data into a custom-built agent using Google Cloud’s Agent Development Kit and Gemini APIs.The system could then be queried for obscure stats, to see if the team could stump Curry and teach him more about his game. Why it matters: For example, it found that his three-point shooting percentage after more than seven dribbles, with a minimum 105 attempts was 40.2%, and how many points Curry generated for his teammates off of screens since 2013: 1,105. Instead of countless hours of manual research, the team got query results in less than a minute. Some queries were so obscure, the team wouldn’t have reached a valid answer without the ability of the agent to analyze the rich data. Learn from us: “Gemini is going to be in my head this year, cause I'm going to be looking at all these details.” – Stephen Curry, Golden State Warriors point guard and 4x NBA champ How CME Group builds a faster, smarter exchange Who: CME Group has evolved from a nineteenth-century commodities exchange into one of the most advanced financial market infrastructures in the world. To support real-time trading and risk management at a global scale, the company launched a strategic partnership with Google Cloud. What they did: By migrating to Cloud SQL and adopting AI-powered insights, CME Group empowered developers, paid down technical debt, and unlocked new opportunities for data-driven innovation across financial markets. Why it matters: Cloud SQL has given CME a foundation for increased developer and team agility. Fewer performance issues mean more time focused on innovation: expanding CME’s analytics capabilities, accelerating AI initiatives, and exploring new ways to commercialize data responsibly. When teams stopped chasing outages, they unlocked more time to take bigger bets and build the future. Learn from us: “With Cloud SQL, we’ve found a way to keep our data layer as fast and dependable as the markets we serve. Cloud SQL gives our teams real-time visibility into what’s happening inside the database. When an application slows, we can identify the root cause in minutes instead of hours. Those insights are built into the platform, which means we don’t need custom tooling or manual analysis to keep operations steady.” – Kristofer Shane Sikora, Executive Director, Cloud Data Engineering, CME Group AppOrchid’s multi-agent system for superior business forecasting Who: App Orchid is an enterprise AI builder and a leader in making data actionable with AI, with a mission to make AI a force for good. Their goal is to empower every employee with trusted, understandable, and accessible data. What they did: The business forecasting agent is actually built on the foundation of two powerful, specialized AI agents: a prediction agent built by Google Cloud and App Orchid’s Data Agent offering. These agents work in concert to solve complex business problems, acting as complementary specialists. App Orchid’s agent possesses unparalleled understanding of an enterprise's past and present, while Google’s agent brings world-class capabilities in predicting the future. Why it matters: Adopting a multi-agent approach provides clear, tangible advantages that directly address the forecasting problems that often plague businesses, including improved accuracy; increased operational efficiency; faster insights; and reduced costs and increased revenue; and greater agility and adaptability. Neither of the underlying agents could achieve these results on their own, but working together, this agent is more than the sum of its subagents. Learn from us: “As the agentic era gets underway, it is evolving quickly. Our multi-agent approach demonstrates both how true agentic systems are most successful when multiple agents are at play, and the importance of finding strong partners with distinct capabilities to help build and assemble these agentic systems.” – Brian Mills, Director, Enterprise AI, Google Cloud & Taka Shinagawa, Gen AI Field Solution Architect, Google Cloud How Mattel uses AI for real-time product updates Who: Virgin Media O2 is one of Europe’s largest telecommunications and media providers, with 45.8 million broadband, mobile, phone, and home subscribers across the UK. To build AI products that are adaptable and data-driven, they needed a decentralized system that internal customers could count on for clean, reliable data. What they did: To improve its understanding of consumer sentiment, Mattel developed an AI-powered feedback classification system, which can analyze millions of customer interactions from a diverse range of sources in a matter of seconds. At its core, the system relies on BigQuery for storing and efficiently processing its massive customer datasets and then utilizes Vertex AI and Google’s multimodal Gemini models to refine and train the sophisticated consumer feedback model. Why it matters: Already, the new AI-powered system has delivered significant wins, delivering a staggering 100x boost in data processing capacity and reducing analysis times from a month to a single minute. By automating the analysis of many processes, analysts are now freed from the noise of everyday tasks, enabling them to focus on deeper research across the company’s iconic portfolio brand. Learn from us: “Our big motto is ‘From months to minutes,’ but it’s real. We were literally spending months-worth of analysis and just getting data into the place that an analyst could tally up all the sentiment — and now it’s just at our fingertips.” – Shaun Applegate, Director of Product Quality Analytics, Mattel Virgin Media O2 uses data contracts for scalable AI products Who: Virgin Media O2 is one of Europe’s largest telecommunications and media providers, with 45.8 million broadband, mobile, phone, and home subscribers across the UK. To build AI products that are adaptable and data-driven, they needed a decentralized system that internal customers could count on for clean, reliable data. What they did: New decentralized data contracts, built with Dataplex, serve as the data quality and assurance layer for VMO2’s data products; these ensure every dataset they publish is reliable, documented, and ready for consumption. Defined at the asset level, such as individual BigQuery tables or Google Cloud Storage buckets, data contracts are redefining how VMO2 manage and share data, enabling the creation of trusted and scalable AI products across their data mesh. Why it matters: The power of this approach lies in moving beyond static documentation. Because they are machine-readable, data contracts become living guarantees with continuous enforcement and real-time validation directly within data pipelines. This proactive monitoring allows teams to detect schema changes or SLA breaches early, transforming data quality from a reactive fix into a scalable, automated mechanism. Learn from us: “By operationalizing trust through data contracts, we are fostering a culture of shared responsibility and data-first thinking. This federated model does more than simply fix pipelines; it builds the trusted foundation needed to scale next-generation AI. It ensures that the resilient AI tools empowering our teams are built on data that is reliable, consistent, and well-defined.” – Chandu Bhuman, Head of Data Strategy, Cloud & Engineering, Virgin Media O2 & Dženan Softić, Data & AI Architect, Google Cloud Inside Mercado Libre's multi-faceted Spanner architecture Who: Mercado Libre, an e-commerce and fintech pioneer across Latin America, operates at a staggering scale, demanding an infrastructure that's not just resilient and scalable, but also a catalyst for rapid innovation. What they did: At the heart of Mercado Libre's strategy is Fury, an in-house middleware platform designed to abstract away the complexities of various backend technologies, providing developers with standardized, simplified interfaces to build applications. Spanner provides Fury with an always-on, globally consistent, multi-model database with virtually unlimited scale. By designating Spanner as a choice within Fury, Mercado Libre ensures that applications built on the platform using Spanner stay consistent globally, scale without breaking, and rarely go down. Why it matters: The strategic adoption of Spanner, amplified by internal platforms like Fury and sophisticated data workflows, has yielded significant benefits, including: significant cost savings and low total cost of ownership; business impact and agility for developers; and low operational overhead thanks to automation. Learn from us: “Mercado Libre's adoption of Spanner demonstrates how to use a powerful, globally consistent database not just for its core capabilities, but as a strategic enabler for developer productivity, operational efficiency, advanced analytics, and future AI ambitions.” – Pablo Leopoldo Arrojo, Software Technical Leader, Mercado Libre Ericsson achieves data integrity and superior governance with Dataplex Who: Mercado Libre, an e-commerce and fintech pioneer across Latin America, operates at a staggering scale, demanding an infrastructure that's not just resilient and scalable, but also a catalyst for rapid innovation. What they did: To power the future of its autonomous network operations and deliver on its strategic priorities, Ericsson's Managed Services, which operates a global network of more than 710,000 sites, has been on a transformative data journey with governance at the center of its strategy.Ericsson moved from foundational practices to a sophisticated, business-enabling data governance framework using the Dataplex Universal Catalog — turning data from a simple resource into a strategic asset. Why it matters: With Dataplex as the governance foundation, Ericsson began implementing the core pillars of its governance program, moving from manual processes to an automated, intelligent data fabric. More specifically, Ericsson established a unified business vocabulary within Dataplex, which helped eliminate ambiguity and ensure their teams — from data scientists to data analysts — were speaking the same language. Learn from us: “Governance is a value enabler, not a blocker. A modern data governance program should focus on business enablement first, driving value and innovation in order to complement policies, rules and risk management. Also remember this work is a journey, not a destination. Be prepared to fail fast, learn, and adapt. The landscape is constantly changing at breakneck speed.” – William McCann Murphy, Head of Data Authority, Ericsson & Akanksha Bhagwanani, EMEA Data Analytics Solution Lead, Google Cloud

The White House's Genesis Mission has set a bold ambition for our nation: to double our scientific productivity within the decade and harness artificial intelligence (AI) to accelerate the pace of discovery. This requires a profound transformation in our national scientific enterprise, one that seamlessly integrates high-performance computing, world-class experimental facilities, and AI. The challenge is no longer generating exabytes of exquisite data from experiments and simulations, but rather curating and exploring it using AI to accelerate the discoveries hidden within. Through our Genesis Mission partnership with the Department of Energy (DOE), Google is committed to powering this new era of federally-funded scientific discovery with the necessary tools and platforms. State-of-the-art reasoning for science The National Labs can take advantage of Gemini for Government—a secure platform with an accredited interface that provides scaled access to a new class of agentic tools designed to augment the scientific process. This includes access to the full capabilities of Gemini, our most powerful and general-purpose AI model. Its native multimodal reasoning operates across the diverse data types of modern science. This means researchers can ask questions in natural language to generate insights grounded in selected sources—from technical reports, code, and images, to a library of enterprise applications, and even organizational and scientific datasets. In addition to the Gemini for Government platform, the National Labs will have access to several Google technologies that support their mission. Today, Google DeepMind announced an accelerated access program for all 17 National Labs, beginning with AI co-scientist—a multi-agent virtual collaborator built on Gemini that can accelerate hypothesis development from years to days—with plans to expand to other frontier AI tools in 2026. Google Cloud provides the secure foundation to bring these innovations to the public sector. By making these capabilities commercially available through our cloud infrastructure, we are ensuring that the latest frontier AI models and tools from Google DeepMind are accessible for the mission-critical work of our National Labs. Accelerating the research cycle with autonomous workflows Gemini for Government brings together the best of Google accredited cloud services, industry-leading Gemini models, and agentic solutions. The platform is engineered to enable autonomous workflows that orchestrate complex tasks. A prime example is Deep Research, which can traverse decades of scientific literature and experimental databases to identify previously unseen connections across different research initiatives or flag contradictory findings that warrant new investigation. By automating complex computational tasks, like managing large-scale simulation ensembles or orchestrating analysis pipelines across hybrid cloud resources, scientists can dramatically accelerate the ‘design-build-test-learn’ cycle, freeing up valuable time for the creative thinking that drives scientific breakthroughs. To ensure agencies can easily leverage these advanced capabilities—including the DOE and its National Laboratories—Gemini for Government is available under the same standard terms and pricing already established for all federal agencies through the General Services Administration's OneGov Strategy. This streamlined access enables National Labs to quickly deploy an AI-powered backbone for their most complex, multi-lab research initiatives. A secure fabric for big team science The future of AI-enabled research requires interconnected experimental facilities, data repositories, and computing infrastructure stewarded by the National Labs. Gemini for Government provides a secure, federated foundation required to reimagine "Big Team Science," creating a seamless fabric connecting the entire DOE complex. AI models and tools in this integrated environment empower researchers to weave together disparate datasets from the field to the benchtop, and combine observations with models, revealing more insights across vast temporal and spatial scales. Ultimately, this transformation can change the nature of discovery, creating a frictionless environment where AI manages complex workflows, uncovers hidden insights, and acts as a true creative research partner to those at our National Labs. Learn more about Gemini for Government by registering for Google Public Sector Summit On-Demand. Ready to discuss how Gemini for Government can address your organization’s needs? Please reach out to our Google Public Sector team at geminiforgov@google.com.

You've built a powerful AI agent. It works on your local machine, it's intelligent, and it's ready to meet the world. Now, how do you take this agent from a script on your laptop to a secure, scalable, and reliable application in production? On Google Cloud, there are multiple paths to deployment, each offering a different developer experience. If you are looking for a detailed architectural comparison to help you choose between Cloud Run, Google Kubernetes Engine (GKE), and Vertex AI Agent Engine, you can start by reading Choosing the Right Deployment Path for Your Google ADK Agents. Ready to build? As part of our Production-Ready AI on Google Cloud Learning Path, we've created three distinct hands-on labs to help you experience these deployment options for yourself. The Managed Solution: Vertex AI Agent Engine For teams seeking the simplest path to production, Vertex AI Agent Engine removes the need to manage web servers or containers entirely. It provides an opinionated environment optimized for python agents, where you define the agent's logic, and the platform handles the execution, memory, and tool invocation. aside_block <ListValue: [StructValue([('title', 'Start the lab!'), ('body', <wagtail.rich_text.RichText object at 0x7fa3d4bc4910>), ('btn_text', ''), ('href', ''), ('image', None)])]> The Serverless Experience: Cloud Run For teams that want the flexibility of containers without the operational overhead, Cloud Run abstracts away the infrastructure, allowing you to deploy your agent as a container that automatically scales up when busy and down to zero when quiet. This path is particularly powerful if you need to build in languages other than Python, use custom frameworks, or integrate your agent into existing declarative CI/CD pipelines. aside_block <ListValue: [StructValue([('title', 'Start the lab!'), ('body', <wagtail.rich_text.RichText object at 0x7fa3d4bc4370>), ('btn_text', ''), ('href', ''), ('image', None)])]> The Orchestrated Experience: Google Kubernetes Engine (GKE) For teams that need precise configuration over their environment, GKE is designed to manage that complexity. This path shows you how an AI agent functions not just as a script, but as a microservice within a broader orchestrated cluster. aside_block <ListValue: [StructValue([('title', 'Start the lab!'), ('body', <wagtail.rich_text.RichText object at 0x7fa3d4bc4310>), ('btn_text', ''), ('href', ''), ('image', None)])]> Your Path to Production Whether you are looking for serverless speed, orchestrated control, or a fully managed runtime, these labs provide the blueprint to get you there. These labs are part of the Deploying Agents module in our official Production-Ready AI with Google Cloud program. Explore the full curriculum for more content that will help you bridge the gap from a promising prototype to a production-grade AI application. Share your progress and connect with others on the journey using the hashtag #ProductionReadyAI. Happy learning!

In computing's early days of the 1940s, mathematicians discovered a flawed assumption about the behavior of round-off errors. Instead of canceling out, fixed-point arithmetic accumulated errors, compromising the accuracy of calculations. A few years later, "random round-off" was proposed, which would round up or down based on a random probability proportional to the remainder. In today's age of generative AI, we face a new numerical challenge. To overcome memory bottlenecks, the industry is shifting to lower precision formats like FP8 and emerging 4-bit standards. However, training in low precision is fragile. Standard rounding destroys the tiny gradient updates driving learning, causing model training to stagnate. That same technique from the 1950s, now known as stochastic rounding, is allowing us to train massive models without losing the signal. In this article, you'll learn how frameworks like JAX and Qwix apply this technique on modern Google Cloud hardware to make low-precision training possible. When Gradients Vanish The challenge in low-precision training is vanishing updates. This occurs when small gradient updates are systematically rounded to zero by "round to nearest" or RTN arithmetic. For example, if a large weight is 100.0 and the learning update is 0.001, a low-precision format may register 100.001 as identical to 100.0. The update effectively vanishes, causing learning to stall. Let's consider the analogy of a digital swimming pool that only records the water level in whole gallons. If you add a teaspoon of water, the system rounds the new total back down to the nearest gallon. This effectively deletes your addition. Even if you pour in a billion teaspoons one by one, the recorded water level never rises. Precision through Probability Stochastic rounding, or SR for short, solves this by replacing deterministic rounding rules with probability. For example, instead of always rounding 1.4 down to 1, SR rounds it to 1 with 60% probability and 2 with 40% probability. Mathematically, for a value x in the interval [⌊x⌋,⌊x⌋+1], the definition is: The defining property is that SR is unbiased in expectation: Stochastic Rounding: E[SR(x)] = x Round-to-Nearest: E[RTN(x)] ≠ x To see the difference, look at our 1.4 example again. RTN is deterministic: it outputs 1 every single time. The variance is 0. It is stable, but consistently wrong. SR, however, produces a noisy stream like 1, 1, 2, 1, 2.... The average is correct (1.4), but the individual values fluctuate. We can quantify the "cost" of zero bias with the variance formula: Var(SR(x))=p(1-p) where p=x-⌊x⌋ In contrast, RTN has zero variance, but suffers from fast error accumulation. In a sum of N operations, RTN's systematic error can grow linearly (O(N)). If you consistently round down by a tiny amount, those errors stack up fast. SR behaves differently. Because the errors are random and unbiased, they tend to cancel each other out. This "random walk" means the total error grows only as the square root of the number of operations O(√N). While stochastic rounding introduces noise, the tradeoff can often be benign. In deep learning, this added variance often acts as a form of implicit regularization, similar to dropout or normalization, helping the model escape shallow local minima and generalize better. Implementing on Google Cloud Google Cloud supports stochastic rounding through its latest generation of AI accelerators, including Cloud TPUs and NVIDIA Blackwell GPUs. These accelerators can also be used in AI-optimized Google Kubernetes Engine clusters. Native Support on TPUs Google's TPU architecture includes native hardware support for stochastic rounding in the Matrix Multiply Unit (MXU). This allows you to train in lower-precision formats like INT4, INT8 and FP8 without meaningful degradation of model performance. You can use Google's Qwix library, a quantization toolkit for JAX that supports both training (QAT) and post-training quantization (PTQ). Here is how you might configure it to quantize a model in INT8, explicitly enabling stochastic rounding for the backward pass to prevent vanishing updates: code_block <ListValue: [StructValue([('code', "import qwix\r\n\r\n# Define quantization rules selecting which layers to compress\r\nrules = [\r\n qwix.QtRule(\r\n module_path='.*',\r\n weight_qtype='int8',\r\n act_qtype='int8',\r\n bwd_qtype='int8', # Quantize gradients\r\n bwd_stochastic_rounding='uniform', # Enable SR for gradients\r\n )\r\n]\r\n\r\n# Apply Quantization Aware Training (QAT) rules\r\nmodel = qwix.quantize_model(model, qwix.QtProvider(rules))"), ('language', 'lang-py'), ('caption', <wagtail.rich_text.RichText object at 0x7fa3e4752eb0>)])]> Qwix abstracts the complexity of low-level hardware instructions, allowing you to inject quantization logic directly into your model's graph with a simple configuration. NVIDIA Blackwell & A4X VMs The story is similar if you are using NVIDIA GPUs on Google Cloud. You can deploy A4X VMs, the industry's first cloud instance powered by the NVIDIA GB200 NVL72 system. These VMs connect 72 Blackwell GPUs into a single supercomputing unit, the AI Hypercomputer. Blackwell introduces native hardware support for NVFP4, a 4-bit floating-point format that utilizes a block scaling strategy. To preserve accuracy, the NVFP4BlockScaling recipe automatically applies stochastic rounding to gradients to avoid bias, along with other advanced scaling techniques. When you wrap your layers in te.autocast with this recipe, the library engages these modes for the backward pass: code_block <ListValue: [StructValue([('code', 'import jax\r\nimport transformer_engine.jax as te\r\nfrom transformer_engine.common.recipe import NVFP4BlockScaling\r\n\r\nkey = jax.random.key(0)\r\nx = jax.random.normal(key, (16, 128, 768))\r\nmodel = te.flax.DenseGeneral(features=768)\r\nparams = model.init(key, x)\r\n\r\ndef loss_fn(params, x):\r\n # NVFP4BlockScaling enables stochastic rounding by default\r\n with te.autocast(recipe=NVFP4BlockScaling()):\r\n output = model.apply(params, x)\r\n return output.mean()\r\n\r\nloss, grads = jax.value_and_grad(loss_fn)(params, x)'), ('language', 'lang-py'), ('caption', <wagtail.rich_text.RichText object at 0x7fa3e4752340>)])]> By simply entering this context manager, the A4X's GB200 GPUs perform matrix multiplications in 4-bit precision while using stochastic rounding for the backward pass, delivering up to 4x higher training performance than previous generations without compromising convergence. Best Practices for Production To effectively implement SR in production, first remember that stochastic rounding is designed for training only. Because it is non-deterministic, you should stick to standard Round-to-Nearest for inference workloads where consistent outputs are required. Second, use SR as a tool for debugging divergence. If your low-precision training is unstable, check your gradient norms. If they are vanishing, enabling SR may help, while exploding gradients suggest problems elsewhere. Finally, manage reproducibility carefully. Since SR relies on random number generation, bit-wise reproducibility is more challenging. Always set a global random seed, for example, using jax.random.key(0), to ensure that your training runs exhibit "deterministic randomness," producing the same results each time despite the internal probabilistic operations. Stochastic rounding transforms the noise of low-precision arithmetic into the signal of learning. Whether you are pushing the boundaries with A4X VMs or Ironwood TPUs, this 1950's numerical method is the key to unlocking the next generation of AI performance. Connect on LinkedIn, X, and Bluesky to continue the discussion about the past, present, and future of AI infrastructure.

In the latest episode of the Agent Factory, Mofi Rahman and I had the pleasure of hosting, Brandon Royal, the PM working on agentic workloads on GKE. We dove deep into the critical questions around the nuances of choosing the right agent runtime, the power of GKE for agents, and the essential security measures needed for intelligent agents to run code. This post guides you through the key ideas from our conversation. Use it to quickly recap topics or dive deeper into specific segments with links and timestamps. Why GKE for Agents? Timestamp: 01:49  We kicked off our discussion by tackling a fundamental question: why choose GKE as your agent runtime when serverless options like Cloud Run or fully managed solutions like Agent Engine exist? Brandon explained that the decision often boils down to control versus convenience. While serverless options are perfectly adequate for basic agents, the flexibility and governance capabilities of Kubernetes and GKE become indispensable in high-scale scenarios involving hundreds or thousands of agents. GKE truly shines when you need granular control over your agent deployments. ADK on GKE Timestamp: 06:58 We've discussed the Agent Development Kit (ADK) in previous episodes, and Mofi highlighted to us how seamlessly it integrates with GKE and even showed a demo with the agent he built. ADK provides the framework for building the agent's logic, traces, and tools, while GKE provides the robust hosting environment. You can containerize your ADK agent, push it to Google Artifact Registry, and deploy it to GKE in minutes, transforming a local prototype into a globally accessible service. The Sandbox problem Timestamp: 15:20 As agents become more sophisticated and capable of writing and executing code, a critical security concern emerges: the risk of untrusted, LLM-generated code. Brandon emphasized that while code execution is vital for high-performance agents and deterministic behavior, it also introduces significant risks in multi-tenant systems. This led us to the concept of a "sandbox." What is a Sandbox? Timestamp: 19:18 For those less familiar with security engineering, Brandon clarified that a sandbox provides kernel and network isolation. Mofi further elaborated, explaining that agents often need to execute scripts (e.g., Python for data analysis). Without a sandbox, a hallucinating or prompt-injected model could potentially delete databases or steal secrets if allowed to run code directly on the main server. A sandbox creates a safe, isolated environment where such code can run without harming other systems. Agent Sandbox on GKE Demo Timestamp: 20:25 So, how do we build this "high fence" on Kubernetes? Brandon introduced the Agent Sandbox on Kubernetes, which leverages technologies like gVisor, an application kernel sandbox. When an agent needs to execute code, GKE dynamically provisions a completely isolated pod. This pod operates with its own kernel, network, and file system, effectively trapping any malicious code within the gVisor bubble.  Mofi walked us through a compelling demo of the Agent Sandbox in action.We observed an ADK agent being given a task requiring code execution. As the agent initiated code execution, GKE dynamically provisioned a new pod, visibly labeled as "sandbox-executor," demonstrating the real-time isolation. Brandon highlighted that this pod is configured with strict network policies, further enhancing security. The Future: Pod Snapshots Timestamp: 29:39 While the Agent Sandbox offers incredible security, the latency of spinning up a new pod for every task is a concern. Mofi demoed the game-changing solution: Pod Snapshots. This technology allows us to save their state of running sandboxes and then near-instantly restore them when an agent needs them. Brandon noted that this reduces startup times from minutes to seconds, revolutionizing real-time agentic workflows on GKE. Conclusion It's incredible to see how GKE isn't just hosting agents; it's actively protecting them and making them faster.  Your turn to build Ready to put these concepts into practice? Dive into the full episode to see the demos in action and explore how GKE can supercharge your agentic workloads. Learn how to deploy an ADK agent to Google Kubernetes Engine and how to get your run agent to run code safely using the GKE agent Sandbox.   Connect with us Shir Meir Lador → LinkedIn, X Mofi Rahman → LinkedIn Brandon Royal → LinkedIn

Welcome to the second Cloud CISO Perspectives for December 2025. Today, Google Cloud’s Nick Godfrey, senior director, and Anton Chuvakin, security advisor, look back at the year that was. As with all Cloud CISO Perspectives, the contents of this newsletter are posted to the Google Cloud blog. If you’re reading this on the website and you’d like to receive the email version, you can subscribe here. aside_block <ListValue: [StructValue([('title', 'Get vital board insights with Google Cloud'), ('body', <wagtail.rich_text.RichText object at 0x7fa3e72f5790>), ('btn_text', 'Visit the hub'), ('href', 'https://cloud.google.com/solutions/security/board-of-directors?utm_source=cloud_sfdc&utm_medium=email&utm_campaign=FY24-Q2-global-PROD941-physicalevent-er-CEG_Boardroom_Summit&utm_content=-&utm_term=-'), ('image', <GAEImage: GCAT-replacement-logo-A>)])]> 2025 in review: Highlighting cloud security and evolving AI By Nick Godfrey, senior director, and Anton Chuvakin, security advisor, Office of the CISO Nick Godfrey, senior director, Office of the CISO Cybersecurity is facing a unique moment, where AI-enhanced threat intelligence, products, and services have begun to give defenders an advantage over the threats they face that had proven elusive — until now. However, threat actors have also begun to take advantage of AI in ways that have moved towards a wider use of tools. At Google Cloud, we continue to strive towards our goals of bringing simplicity, streamlining operations, and enhancing efficiency and effectiveness for security essentials. AI is now part of that essential security approach, both building AI securely and using AI to boost defenders. Anton Chuvakin, security advisor, Office of the CISO Looking back at 2025, we’re sharing our top stories across five vital areas of development in cybersecurity: securing cloud, securing AI, AI-enabled defense, threat intelligence, and building the most trusted cloud. Securing cloud This year reinforced the importance of cloud security fundamentals. Cybersecurity risks continue to accelerate with the number and severity of breaches continuing to grow, and more organizations are turning to multi-cloud and hybrid solutions that introduce their own complex management challenges. Google announces agreement to acquire Wiz Responding to CVE-2025-55182: Secure your React and Next.js workloads Google advances sovereignty, choice, and security in the cloud Prove your expertise with our new SecOps engineer certification Project Shield blocked a massive recent DDoS attack. Here’s how Secure cloud. Insecure use. (And what you can do about it) How Google Does It: Modernizing threat detection Securing AI 2025 was a crucial year as we continued our efforts to build AI securely — and to encourage others to do so, too. From AI governance to building agents securely, we wanted to give our customers the tools they need to secure their AI supply chain and tools. Boards should be ‘bilingual’ in AI, security to gain advantage 5 tips for secure AI success Introducing AI Protection: Security for the AI era Google guidance on securing your AI supply chain How Google secures AI agents AI agent security: How to protect digital sidekicks (and your business) AI-enabled defense We have seen some incredible strides towards empowering defenders with AI this year. As defenders guide others on how to secure their use of AI, we must ensure that we also use AI to support stronger defensive action. Our Big Sleep agent makes a big leap 3 new ways to use AI as your security sidekick How Google Does It: Building AI agents for cybersecurity and defense AI as a strategic imperative to manage risk Beyond the hype: Analyzing new data on ROI of AI in security The dawn of agentic AI in security operations at RSAC 2025 Introducing the Agentic SOC Workshops for security professionals Threat intelligence As defenders have made significant advances in using AI to boost their efforts this year, government-backed threat actors and cybercriminals have been trying to do the same. At Google, we strongly believe in the power of threat intelligence to enhance defender abilities to respond to critical threats faster and more efficiently. The ultimate insider threat: North Korean IT workers Recent advances in how threat actors use AI tools New AI, cybercrime reports underscore need for security best practices How CISOs and boards can help fight cyber-enabled fraud How Google Does It: Using threat intelligence to uncover and track cybercrime How to build a best-practice Cyber Threat Intelligence program Building the most trusted cloud We continued to enhance our security capabilities and controls on our cloud platform to help organizations secure their cloud environments and address evolving policy, compliance, and business objectives. Announcing the Google Unified Security Recommended program Next ‘25: Driving secure innovation with Google Unified Security Security Summit 2025: Enabling defenders and securing AI innovation Disrupt ransomware with AI in Google Drive Enabling a safe agentic web with reCAPTCHA Mastering secure AI on Google Cloud: A practical guide for enterprises Our forecast for 2026 As security professionals, we know that threat actors will continue to innovate to achieve their mission objectives. To help defenders proactively prepare for the coming year, we publish our annual forecast report with insights from across Google. We look forward to sharing more insights to help organizations strengthen their security posture in the new year. For more leadership guidance from Google Cloud experts, please visit our CISO Insights hub. aside_block <ListValue: [StructValue([('title', 'Learn something new'), ('body', <wagtail.rich_text.RichText object at 0x7fa3e72f59a0>), ('btn_text', 'Watch now'), ('href', 'https://youtu.be/7gNyN3fBn00?si=Lgrxi9908i-IvAHP'), ('image', <GAEImage: Cloud-CISO-Perspectives-logo-A>)])]> In case you missed it Here are the latest updates, products, services, and resources from our security teams so far this month: How Google Does It: Collecting and analyzing cloud forensics: Here’s how Google’s Incident Management and Digital Forensics team gathers and analyzes digital evidence. Read more. When securing Web3, remember your Web2 fundamentals: As Web3 matures, the stakes continue to rise. For Web3 to thrive, security should expand beyond the blockchain to protect operational infrastructure. Here’s how. Read more. How Mandiant can help test and strengthen your cyber resilience: To help teams better prepare for actual incidents, we developed ThreatSpace, a cyber proving ground with all the digital noise of real employee activities. Read more. Exploiting agency of autonomous AI agents with task injection: Learn what a task injection attack is, how it differs from prompt injection, and how it is particularly relevant to AI agents designed for a wide range of actions and tasks, such as computer-use agents. Read more. Please visit the Google Cloud blog for more security stories published this month. aside_block <ListValue: [StructValue([('title', 'Join the Google Cloud CISO Community'), ('body', <wagtail.rich_text.RichText object at 0x7fa3e72f5760>), ('btn_text', 'Learn more'), ('href', 'https://rsvp.withgoogle.com/events/ciso-community-interest?utm_source=cgc-blog&utm_medium=blog&utm_campaign=2024-cloud-ciso-newsletter-events-ref&utm_content=-&utm_term=-'), ('image', <GAEImage: GCAT-replacement-logo-A>)])]> Threat Intelligence news How threat actors are exploiting React2Shell: Shortly after CVE-2025-55182 was disclosed, Google Threat Intelligence Group (GTIG) began observing widespread exploitation across many threat clusters, from opportunistic cybercrime actors to suspected espionage groups. Here’s what GTIG has observed so far. Read more. Intellexa’s prolific zero-day exploits continue: Despite extensive scrutiny and public reporting, commercial surveillance vendors such as Intellexa continue to operate unimpeded. Known for its “Predator” spyware, new GTIG analysis shows that Intellexa is evading restrictions and thriving. Read more. APT24's pivot to multi-vector attacks: GTIG is tracking a long-running and adaptive cyber espionage campaign by APT24, a People's Republic of China (PRC)-nexus threat actor that has been deploying BADAUDIO over the past three years. Here’s our analysis of the campaign and malware, and how defenders can detect and mitigate this persistent threat. Read more. Please visit the Google Cloud blog for more threat intelligence stories published this month. Now hear this: Podcasts from Google Cloud Bruce Schneier on the AI offense-defense balance: From rewiring democracy to hacking trust, Bruce Schneier discusses the impact of AI on society with hosts Anton Chuvakin and Tim Peacock. Hear his take on whether it will help support liberal democracy more, or boost the forces of corruption, illiberalism, and authoritarianism. Listen here. The truth about autonomous AI hacking: Heather Adkins, Google’s Security Engineering vice-president, separates the hype from the hazards of autonomous AI hacking, with Anton and Tim. Listen here. Escaping 1990s vulnerability management: Caleb Hoch, consulting manager for security transformations, Mandiant, discusses with Anton and Tim how vulnerability management has evolved beyond basic scanning and reporting, and the biggest gaps between modern practices and what organizations are actually doing. Listen here. Adopting a dual offensive-defensive mindset: Betty DeVita, private and public board director and fintech advisor, shares her take on how boards can take on an offensive and defensive approach to cybersecurity for their organizations. Listen here. To have our Cloud CISO Perspectives post delivered twice a month to your inbox, sign up for our newsletter. We’ll be back in a few weeks with more security-related updates from Google Cloud.

Data teams seem to be constantly balancing the need for governed, trusted metrics with business needs for agility and ad-hoc analysis. To help bridge the gap between managed reporting and rapid data exploration, we are introducing several new features in Looker, to expand users’ self-service capabilities. These updates allow individuals to analyze local data alongside governed models, organize complex dashboards more effectively, and align the look and feel of their analytics with corporate branding, all within the Looker platform. Analyze ad-hoc data with Looker self-service Explores Valuable data often exists outside of the primary database — whether in budget spreadsheets, sales lists, or ad-hoc research files. With self-service Explores, now in Preview, users can upload CSV and spreadsheet-based data using a drag-and-drop interface directly within Looker. This feature allows users to combine local files with fully modeled Looker data to test new theories and enrich insights. Once uploaded, users can visually add new measures and dimensions to their self-service Explores, customize them, and share the results via dashboards and Looks. Uploading a CSV file and creating a new self-service Explore in just a few clicks To maintain governance, administrators retain oversight regarding which files are uploaded to the Looker instance and who has permission to perform uploads. Additionally, we have introduced a new content certification flow, which makes it easier to signal which content is the vetted, trusted source of truth, ensuring users can distinguish between ad-hoc experiments and certified data. Certifying a self-service Explore Upload data and content certification are available in Public Preview as of Looker 25.20. Deliver clearer, cohesive data stories with tabbed dashboards The new tabbed dashboard feature helps dashboard editors organize complex information into logical narratives, moving away from dense, single-page views. Editors can now streamline content creation with controls for adding, renaming, and reordering tabs. For the viewer, the experience is designed to be seamless. Filters automatically pass values across the entire dashboard, while each tab displays only the filters relevant to the current view, reducing visual clutter. Users can share unique URLs for specific tabs and schedule or download the comprehensive multi-tab dashboard as a single PDF document. Navigating between tabs on a multi-tab dashboard This feature is currently available in preview. Apply custom styling to dashboards Matching internal dashboards to company branding can help create a familiar data experience and increase user engagement. We are announcing the Public Preview of internal dashboard theming, which allows creators to apply custom changes to tile styles, colors, fonts, and formatting directly to dashboards consumed inside the Looker application. Applying custom theming for internal dashboards With this feature, you can save, share, and apply pre-configured themes to ensure consistency. Users with permission to manage internal themes can create new templates for existing dashboards or select a default theme to apply across the entire instance. You can enable Internal dashboard theming today on the Admin > Labs page. Enabling the preview for internal dashboard theming Get started These new self-service capabilities in Looker are designed to help you and all users in your organization get more value out of your data by improving presentation flexibility and quality. Try self-service Explores and internal dashboard themes for yourself today and let us know your feedback.

In the AI era, when one year can feel like 10, you’re forgiven for forgetting what happened last month, much less what happened all the way back in January. To jog your memory, we pulled the readership data for top product and company news of 2025. And because we publish a lot of great thought leadership and customer stories, we pulled that data too. Long story short: the most popular stories largely mapped to our biggest announcements. But not always — there were more than a few sleeper hits on this year’s list. Read on to relive this huge year, and perhaps discover a few gems that you may have missed.  Building tomorrow, today: 2025 customer AI innovation highlights with Google Cloud January 2025 started strong with important new virtual machine offerings, foundational AI tooling, and tools for both Kubernetes and data professionals. We also launched our "How Google Does It" series, looking at the internal systems and engineering principles behind how we run a modern threat-detection pipeline. We showed developers how to get started with JAX and made AI predictions for the year ahead. Readers were excited to learn about how L’Oréal built its MLOps platform and Deutsche Börse’s pioneering work on cloud-native financial trading. Product news Simplify the developer experience on Kubernetes with KRO Blackwell is here — new A4 VMs powered by NVIDIA B200 now in preview Introducing Vertex AI RAG Engine: Scale your Vertex AI RAG pipeline with confidence Introducing BigQuery metastore, a unified metadata service with Apache Iceberg support C4A, the first Google Axion Processor, now GA with Titanium SSD Thought leadership: How Google Does It: Making threat detection high-quality, scalable, and modern 2025 and the Next Chapter(s) of AI Customer stories How L’Oréal Tech Accelerator built its end-to-end MLOps platform Trading in the Cloud: Lessons from Deutsche Börse Group’s cloud-native trading engine February There are AI products, and then there are products enhanced by AI. This month’s top launch, Gen AI Toolbox for Databases, falls into the latter category. This was also the month readers got serious about learning, with blogs about upskilling, resources, and certifications topping the charts. The fruits of our partnership with Anthropic made an appearance in our best-read list, and engineering leaders detailed Google’s extensive efforts to optimize AI system energy consumption. Execs ate up an opinion piece about how agents will unlock insights into unstructured data (which makes up 90% of enterprises’ information assets), and digested a sobering report on AI and cybercrime. During the Mobile World Congress event, we saw considerable interest in our work with telco leaders like Vodafone Italy and Amdocs. Product and company news: Announcing public beta of Gen AI Toolbox for Databases Get Google Cloud certified in 2025—and see why the latest research says it matters Discover Google Cloud careers and credentials in our new Career Dreamer Announcing Claude 3.7 Sonnet, Anthropic’s first hybrid reasoning model, is available on Vertex AI Thought leadership Designing sustainable AI: A deep dive into TPU efficiency and lifecycle emissions From dark data to bright insights: How AI agents make data simple New AI, cybercrime reports underscore need for security best practices Customer stories Transforming data: How Vodafone Italy modernized its data architecture in the cloud AI-powered network optimization: Unlocking 5G's potential with Amdocs March Back when we announced it, our intent to purchase cybersecurity startup Wiz was Google’s largest deal ever, and the biggest tech deal of the year. We built on that security momentum with the launch of AI Protection. We also spread our wings to the Nordics with a new region, and announced the Gemma 3 open model on Vertex AI. Meanwhile, we explained the threat that North Korean IT workers pose to employers, gave readers a peek under the hood of the Colossus file system, and reminisced about what we’ve learned over 25 years of building data centers. Readers were interested in Levi’s approach to data and weaving it into future AI efforts, and in honor of the GDC Festival of Gaming, our AI partners shared some new perspectives on “living games.” Product and company news Google + Wiz: Strengthening Multicloud Security Announcing AI Protection: Security for the AI era Hej Sverige! Google Cloud launches new region in Sweden Announcing Gemma 3 on Vertex AI Thought leadership The ultimate insider threat: North Korean IT workers Colossus under the hood: How we deliver SSD performance at HDD prices 3 key lessons from 25 years of warehouse scale computing Customer stories Levi's seamless data strategy: How tailor-made AI keeps an icon from getting hemmed in Co-op mode: New partners driving the future of gaming with AI April With April came Google Cloud Next, our flagship annual conference. From Firebase Studio, Ironwood TPUs, and Google Agentspace, to Vertex AI, Cloud WAN, and Gemini 2.5, it’s hard to limit ourselves to just a few stories, there were so many bangers (for the whole list, there’s always the event recap). Meanwhile, our systems team discussed innovations to keep data center infrastructure’s thermal envelope in check. And at the RSA Conference, we unveiled our vision for the agentic security operations center of the future. On the customer front, we highlighted the startups who played a starring role at Next, and took a peek behind the curtain of The Wizard of Oz at Sphere. Product and company news Introducing Firebase Studio and agentic developer tools to build with Gemini Introducing Ironwood TPUs and new innovations in AI Hypercomputer Vertex AI offers new ways to build and manage multi-agent systems Scale enterprise search and agent adoption with Google Agentspace Cloud WAN: Connect your global enterprise with a network built for the AI era Gemini 2.5 brings enhanced reasoning to enterprise use cases The dawn of agentic AI in security operations at RSAC 2025 Thought leadership AI infrastructure is hot. New power distribution and liquid cooling infrastructure can help 3 new ways to use AI as your security sidekick Customer stories Global startups are building the future of AI on Google Cloud The AI magic behind Sphere’s upcoming 'The Wizard of Oz' experience May School was almost out, but readers got back into learning mode to get certified as generative AI leaders. You were also excited about new gen AI media models in Vertex AI, the availability of Anthropic’s Claude Opus 4 and Claude Sonnet 4. We also learned that you’re very excited to use AI to generate SQL code, and about using Cloud Run as a destination for your AI apps. We outlined the steps for building a well-defined data strategy, and showed governments how AI can actually improve their security posture. And on the customer front, we launched our “Cool Stuff Customers Built” round-ups, and ran stories from Formula E and MLB. Google Cloud announces first-of-its-kind generative AI leader certification Expanding Vertex AI with the next wave of generative AI media models Announcing Anthropic’s Claude Opus 4 and Claude Sonnet 4 on Vertex AI Thought leadership Getting AI to write good SQL: Text-to-SQL techniques explained AI deployments made easy: Deploy to Cloud Run from AI Studio or any MCP client Building a data strategy for the AI era How governments can use AI to improve threat detection and reduce cost Customer stories Cool Stuff Customers Built: May Edition Pushing the limits of electric mobility: Formula E's Mountain Recharge Tuning in with AI: How MLB My Daily Story creates truly personalized highlight videos June Up until this point, the promise of generative AI was largely around text and code. The launch of Veo 3 changed all that. Developers writing and deploying AI apps saw the availability of GPUs on Cloud Run as a big win, and we continued our steady drumbeat of Gemini innovation with 2.5 Flash and Flash-Lite. We also shared our thoughts on securing AI agents. And to learn how to actually build these agents, readers turned to stories about Box, the British real estate firm Schroders, and French luxury conglomerate LVMH (home of Louis Vuitton, Channel, Sephora and more). You dream it, Veo creates it: Veo 3 is now available for everyone in public preview on Vertex AI Cloud Run GPUs, now GA, makes running AI workloads easier for everyone Gemini momentum continues with launch of 2.5 Flash-Lite and general availability of 2.5 Flash and Pro on Vertex AI Thought leadership Ask OCTO: Making sense of AI agents Cloud CISO Perspectives: How Google secures AI agents Customer stories The secret to document intelligence: Box builds Enhanced Extract Agents with A2A framework How Schroders built its multi-agent financial analysis research assistant Inside LVMH's perfectly manicured data estate, where luxury AI agents are taking root July Readers took a break from reading about AI to read about network infrastructure — the new Sol transatlantic cable, to be precise. Then it was back to AI: new video generation models in Vertex; a crucial component for building stateful, context-aware agents; and a new toolset for connecting BigQuery data to Agent Development Kit (ADK) and Multi-Cloud Protocol (MCP) environments. Developers cheered the integration between Cloud Run and Docker Compose, and executive audiences enjoyed a listicle on actionable, real-world uses for AI agents. On the security front, we took a back-to-basics approach this month, exploring the persistence of some cloud security problems. And then, back to AI again, with our Big Sleep agent. Readers were also interested in how AI is alleviating record-keeping for nurses at HCA Healthcare, Ulta Beauty’s data warehousing and mobile record keeping initiatives, and how SmarterX migrated from Snowflake to BigQuery. Strengthening network resilience with the Sol transatlantic cable Veo 3 and Veo 3 Fast are now generally available on Vertex AI Announcing Vertex AI Agent Engine Memory Bank available for everyone in preview BigQuery meets ADK & MCP: Accelerate agent development with BigQuery's new first-party toolset From localhost to launch: Simplify AI app deployment with Cloud Run and Docker Compose Thought leadership Secure cloud. Insecure use. (And what you can do about it) Our Big Sleep agent makes a big leap Customer stories How nurses are charting the future of AI at America's largest hospital network, HCA Healthcare Ulta Beauty redefines beauty retail with BigQuery SmarterX’s migration from Snowflake to BigQuery accelerated model building and cut costs in half August AI is compute- and energy-intensive; in a new technical paper, we released concrete numbers about our AI infrastructure’s power consumption. Then people went [nano] bananas for Gemini 2.5 Flash Image on Vertex AI, and developers got a jump on their AI projects with a wealth of technical blueprints to work from. The summer doldrums didn’t stop our security experts from tackling the serious challenge of cyber-enabled fraud. We also took a closer look at the specific agentic tools empowering workers at Wells Fargo, and how Keeta processes 11 million blockchain transactions per second with Spanner. How much energy does Google’s AI use? We did the math Building next-gen visuals with Gemini 2.5 Flash Image (aka nano-banana) on Vertex AI 101+ gen AI use cases with technical blueprints Thought leadership New Threat Horizons report details evolving risks — and defenses How CISOs and boards of directors can help fight cyber-enabled fraud How AI-powered weather forecasting can transform energy operations Customer stories How Wells Fargo is using Google Cloud AI to empower its workforce with agentic tools How Keeta processes 11 million financial transactions per second on the blockchain with Spanner September AI is cool tech, but how do you monetize it? One answer is the Agent Payment Protocol, or AP2. Developers and data scientists preparing for AI flocked to blogs about new Data Cloud offerings, the 2025 DORA Report, and new trainings. Executives took in our thoughts on building an agentic data strategy, and took notes on the best prompts with which to kickstart their AI usage. And because everybody is impacted by the AI era, including business leaders, we explained what it means to be “bilingual” in AI and security. Then, at Google's AI Builders Forum, startups described how Google’s AI, infrastructure, and services are supporting their growth. Not to be left out, enterprises like Target and Mr. Cooper also showed off their AI chops. Powering AI commerce with the new Agent Payments Protocol (AP2) The new data scientist: From analyst to agentic architect Announcing the 2025 DORA Report: State of AI-Assisted Software Development Back to AI school: New Google Cloud training to future-proof your AI skills Thought leadership Building better data platforms, for AI and beyond Boards should be ‘bilingual’ in AI, security to gain advantage A leader’s guide to five essential AI prompts Customer stories How Google Cloud’s AI tech stack powers today’s startups From query to cart: Inside Target’s search bar overhaul with AlloyDB AI How Mr. Cooper assembled a "team" of AI agents to handle complex mortgage questions October Welcome to the Gemini Enterprise era, which brings enhanced security, data control, and advanced agent capabilities to large organizations. To help you prepare, we relaunched a variety of enhancements to our learning platform, and added new commerce and security programs. And while developers versed themselves on the finer points of Veo prompts, we discussed securing the AI supply chain, building AI agents for cybersecurity and defense, and a new vision on economic threat modeling. We partnered with PayPal to enable commerce in AI chats, Germany’s Planck Institute showed how AI can help share deep scientific expertise, and DZ Bank pioneered ways to make blockchain-based finance more reliable. Introducing Gemini Enterprise Google Skills: Your new home for cloud learning Enabling a safe agentic web with reCAPTCHA Partners powering the Gemini Enterprise agent ecosystem Thought leadership The ultimate prompting guide for Veo 3.1 How you can secure your AI supply chain How Google Does It: Building AI agents for cybersecurity and defense Customer stories Introducing an agentic commerce solution for merchants from PayPal and Google Cloud How the Max Planck Institute is sharing expert skills through multimodal agents The oracles of DeFi: How DZ Bank builds trustworthy data feeds for decentralized applications November Whether it was Gemini 3, Nana Banana Pro, or our seventh-generation Ironwood TPUs, this was the month that we gave enterprise customers access to all our latest and greatest AI tech. We also did a deep dive on how we built the largest-ever Kubernetes cluster, clocking in at a massive 130,000 nodes, and we announced a new collaboration with AWS to improve connectivity between clouds. Meanwhile, we updated our findings on the adversarial misuse of AI by threat actors and on the ROI of AI for security, and executives vibed out on our piece about vibe coding. Then, just in time for the holidays, we took a look at how Mattel is using AI tools to revamp its toys, and Waze showed how it uses Memorystore to keep the holiday traffic flowing. Bringing Gemini 3 to Enterprise How Google Does It: Building the largest known Kubernetes cluster, with 130,000 nodes Announcing Nano Banana Pro for every builder and business Announcing Ironwood TPUs General Availability and new Axion VMs to power the age of inference AWS and Google Cloud collaborate to simplify multicloud networking Thought leadership Recent advances in how threat actors use AI tools Beyond the hype: Analyzing new data on ROI of AI in security How vibe coding can help leaders move faster Customer stories Mattel’s game changer: How AI is turning customer feedback into real-time product updates Waze keeps traffic flowing with 1M+ real-time reads per second on Memorystore December The year is winding down, but we still have lots to say. Early returns show that you were interested in how to mitigate the React2Shell vulnerability, support for MCP across Google services, and the early access launch of AlphaEvolve. And let’s not forget Gemini 3 Flash, which is turning heads with its high-level reasoning, plus amazing speed and a flexible cost profile. What does this all mean for you and your future? It’s important to contextualize these technology developments, especially AI. For example, the DORA team put together a guide on how high-performing platform teams can integrate AI capabilities into their workflows, we discussed what it looks like to have an AI-ready workforce, and our Office of the CISO colleagues put out their 2026 cybersecurity predictions. More to the point (guard), you could do like Golden State Warrior Stephen Curry and turn to Gemini to analyze your game, to prepare for the year ahead. We’ll be watching on Christmas Day to see how Steph is faring with Gemini’s advice. Responding to React2Shell (CVE-2025-55182): Secure your React and Next.js workloads Announcing Model Context Protocol (MCP) support for Google services AlphaEvolve on Google Cloud: AI for agentic discovery and optimization Introducing Gemini 3 Flash: Intelligence and speed for enterprises Thought leadership From adoption to impact: Putting the DORA AI Capabilities Model to work Is AI fluency the ingredient or the result of an AI-ready workforce? Our 2026 Cybersecurity Forecast report Customer stories What Stephen Curry learned about his game from a custom Gemini agent The Curry sibling rivalry is going strong And that’s a wrap on 2025! Thanks for reading, and see you next year!

Want to know the latest from Google Cloud? Find it here in one handy location. Check back regularly for our newest updates, announcements, resources, events, learning opportunities, and more.  Tip: Not sure where to find what you’re looking for on the Google Cloud blog? Start here: Google Cloud blog 101: Full list of topics, links, and resources. Dec 15 - Dec 19 Announcing Advanced Governance Capabilities to Vertex AI Agent Builder: Today, with the integration of the Cloud API Registry, we’re excited to bring enhanced tool governance capabilities to Vertex AI Agent Builder. With this latest update, administrators can now manage available tools for developers across your organization directly in Vertex AI Agent Builder Console, and developers can leverage tools managed by the registry with a new ApiRegistry. Following last month's expansion of our Agent Builder platform, we are also introducing new capabilities across the entire agent lifecycle to help developers build faster using new ADK capabilities and visual tools, and scale with high performance through the expansion of Agent Engine services, including the general availability of support for sessions and memory. Read more below. Vertex AI Agent Builder provides the unified platform to manage the entire agent lifecycle, helping you close the gap from prototype to a production-ready agent. To explore these new features, visit the updated Agent Builder documentation and release notes. Single-tenant Cloud HSM is now Generally Available: We’re thrilled to announce the General Availability (GA) of Single-tenant Cloud HSM -  a standards compliant, highly available, and scalable HSM cluster that provides you complete control over your cryptographic keys for highly sensitive workloads in the cloud for general purpose applications. Customers have complete control over their cryptographic keys and the ability to manage their own admin credentials through our gcloud APIs, which establish a cryptographically isolated cluster of dedicated HSM partitions for each customer. Single-tenant Cloud HSM is integrated with Cloud KMS, allowing its use with Customer-Managed Encryption Keys (CMEK). Single-tenant Cloud HSM is available in the following regions: us-central1, us-east4, europe-west1, and europe-west4. Advanced AI, data, and compliance security capabilities are now available to Security Command Center (SCC) Premium pay-as-you-go (PayGo) customers. Previously exclusive to Enterprise and Premium subscriptions, we now offer to PayGo customers the AI Security Dashboard, Data Security Posture Management (DSPM), Compliance Manager, and Security Graph, including Graph Search and Correlated Threats. This update can help you address novel risks from generative AI and autonomous agents by offering integrated, automated protection for both traditional and AI workloads in Google Cloud. Customers can start a 30-day free trial to access the full SCC Premium experience. Dec 8 - Dec 12 Application Design Center is now Generally Available We're excited to announce the General Availability (GA) of Application Design Center, enabling platform teams and developers to streamline cloud application infrastructure design, deployment, and evolution, ensuring security and best practices. This GA launch includes powerful new capabilities such as enterprise-grade governance with public APIs and gcloud CLI support; bring your own Terraform, full compatibility with VPC Service Controls; and simplified onboarding with app-managed project support. To learn more, read the Application Design Center GA launch blog. Apigee Feature Templater Simplifies API Proxy Development for Everyone The new open-source Apigee Feature Templater (AFT) streamlines API proxy authoring by turning complex policies into reusable building blocks called "features." Non-experts can quickly assemble robust proxies—including AI Gateways, security, and data masking—using simple CLI or REST commands. AFT accelerates time-to-market by enabling expert developers to delegate feature creation and empowering a broader team to compose APIs. Read the full release details. Navigating the Industry Shift in Client Authentication for Apigee mTLS An industry policy change is phasing out the Client Authentication Extended Key Usage (EKU) in public certificates, directly impacting server-to-server mTLS for Apigee. This shift forces organizations away from Public CAs to manage their own Private PKI to maintain service continuity by mid-2026. This article presents the two paths: implementing a Private Certificate Authority (Private CA), ideally using Google Cloud Certificate Authority Service (CAS) for immediate mTLS continuity; or modernizing long-term with Demonstrating Proof of Possession (DPoP) for maximum operational efficiency. Read about the two paths to mTLS continuity. Learn how to implement Kubernetes Secrets in Apigee hybrid Apigee hybrid introduces direct, read-only access to custom Kubernetes Secrets within API proxies. This exclusive feature offers a superior way to handle highly sensitive credentials (like private keys and backend passwords) compared to KVMs. Secrets never leave the cluster boundary, ensuring enhanced compliance and security. It enables a clean separation of duties, allowing cluster operators to manage credentials via GitOps workflows while API developers securely reference them using flow variables, without ever viewing the raw sensitive data. Read the full article. Don't let your AI program fail at the final hurdle. Our new guide, Successful Chatbot: 5 Steps from ROI to Rollout, outlines the essential practices for rigorous customer testing and strategic deployment. Learn how to align testing with business goals, define clear evaluation criteria, and drive actionable insights. The post emphasizes that delivering a successful AI program requires more than just domain expertise, highlighting the importance of clear scoping, strategic staffing, and disciplined financial planning. This is crucial for maximizing confidence in your AI's long-term impact, especially in regulated industries like healthcare. Useful Product links - Google Cloud's Vertex AI & Agentspace, Google Cloud's Healthcare API, and, Google Cloud's Natural Language API Apigee Now Supports Model Context Protocol (MCP) for Secure Agentic Tools Google has expanded support for Model Context Protocol (MCP) with the release of fully-managed, remote MCP servers, giving developers worldwide consistent and enterprise-ready access to Google and Google Cloud services. This includes support for MCP in Apigee, which makes it possible for agents to use your secure, governed APIs and custom workflows cataloged in Apigee API hub as tools to complete tasks for end users. With Apigee’s support for MCP, you don’t need to make any changes to your existing APIs, write any code, or deploy and manage any local or remote MCP servers. Apigee uses your existing API specifications and manages the underlying infrastructure and transcoding, so that you can focus on the business logic for your agents. Read the full announcement. Introducing Fully-Managed MCP Servers to Power Agentic AI Google Cloud is announcing fully-managed, remote Model Context Protocol (MCP) servers, enhancing Google’s API infrastructure to provide a unified, enterprise-ready layer for AI agents. This eliminates the burden on developers to install and manage individual MCP servers. Now, AI agents can reliably use trusted Google services like Google Maps, BigQuery, Compute Engine, and GKE as tools to perform complex tasks. This unified approach, managed via Cloud IAM and Apigee API Hub, ensures rigorous governance, security, and observability for all agentic interactions. Read the full announcement. Marketplace Customer Credit Program now available for Marketplace Channel Private Offers Google Cloud's Marketplace Customer Credit Program offers up to 3% in Google Cloud credits when customers purchase an eligible cloud marketplace solution for the first time, whether directly through an ISV or via a chosen channel partner. Learn more. Two-step control plane minor upgrades with rollback safety in Public preview Upgrading a production Kubernetes cluster can be a stressful, high-stakes event. GKE's new two-step control plane minor upgrade with rollback safety gives you a safe window to validate a new minor version before committing, with a simple rollback option if you find any issues. By decoupling control plane binary changes from new API and feature changes, you can easily revert to the previous minor version if issues arise during the validation period. Learn more about this new, safer upgrade process. Google named a Leader in the 2025 IDC MarketScape for Worldwide Hyperscaler Marketplaces IDC Marketscape has positioned Google as a Leader in the 2025 IDC MarketScape for Worldwide Hyperscaler Marketplaces. We believe this recognition underscores our commitment to deliver a cloud marketplace experience that fuels the AI agent economy and accelerates innovation. This achievement reflects our dedication to creating an open and interoperable agentic AI ecosystem for our customers and partners. Learn more. Dec 1 - Dec 5 Unlock the full potential of your data with Object Contexts in Google Cloud Storage This new feature provides a foundation for semantic storage and actions, allowing you to integrate Gemini with GCS objects and enrich your objects in a more intelligent and meaningful way. Learn how to get started with Object Contexts and revolutionize your data workflows. Learn more. Nov 24 - Nov 28 Boost API Security: IP Allowlisting and ML Enhancements for Apigee Abuse Detection To keep your applications safe, it’s critical to detect and block attacks on your APIs as quickly as possible. In the past few months, we’ve made some changes to Advanced API Security’s Abuse Detection feature to make it easier and faster to identify legitimate attacks and take action. Get all the details on Apigee's new IP allowlisting. Apigee AI Gateway Deep Dive on December 11 Join the final Apigee Community Tech Talk of the year for a deep dive into the Apigee AI Gateway. This session provides practical details on integrating, proxying, and converting complex MCP protocol services with traditional REST backends. Learn specific techniques for securing, monitoring, and gaining technical control over MCP backends to meet enterprise-grade governance requirements. Register now Nov 10 - Nov 14 Deploy n8n to Cloud Run With just a few commands, you can deploy n8n to Cloud Run and have it up and running, ready to supercharge your business with AI workflows that can manage spreadsheets, read and draft emails, and more. n8n and Cloud Run are both easy to use and powerful tools that empower developers to do more with AI. Learn more here. GKE Node Memory Swap in Public preview You can now configure swap space on your GKE Standard nodes to provide a crucial buffer against Out-of-Memory (OOM) errors for memory-intensive applications, especially during unexpected usage spikes. Enabling swap can improve workload resilience, reduce pod evictions due to memory pressure, and enhance overall application stability and cost-effectiveness. This feature is currently available in a public preview. Nov 3 - Nov 7 Announcing the Data Engineering Agent Data teams can now automate complex SQL pipeline tasks with the new Data Engineering Agent for BigQuery, available in Preview. This agent simplifies development, maintenance, and troubleshooting, allowing engineers to focus on strategic initiatives. It supports natural language pipeline creation, intelligent modification, and seamless migration from legacy tools. Transform your data engineering workflows today! From Threat Model to TTX: Bringing a New Design Partner to the Table Gain an overview of threat modeling, how threat models can be performed rapidly, and why threat model scenarios make excellent tabletop scenarios - especially for products that are still in development. To get more information about threat modeling or tabletop exercises, check out The Defender’s Advantage or reach out to a Mandiant cybersecurity expert for specialized assistance. Application Monitoring now includes a Topology. Application Monitoring now includes a graphical representation of runtime dependencies (i.e Topology) for your App Hub defined application. This now allows you to quickly understand your app architecture, spot anomalous runtime interactions and resolve issues flagged from alerts quicker. Runtime dependencies are extracted from the OpenTelemetry traces you send to Cloud Trace from your App Hub registered workload. Follow the outline here to register your app and unlock the benefits of Application Monitoring and its newly launched Topology Supercharge AI Agents: Apply Enterprise Governance to GenAI Workflows with Apigee As Generative AI agents move to production, you need control over cost, reliability, and security. A powerful new pattern introduces Apigee as the unified AI Agent Gateway for Large Language Model (LLM) calls. Route agent traffic through Apigee to gain immediate enterprise-grade governance, including dynamic circuit breaking, token consumption quotas, and sensitive data masking. A new Apigee wrapper for the Agent Development Kit (ADK) simplifies implementation. Turn your agents into manageable, secure AI products. Read the full article and explore the new pattern. Oct 20 - Oct 24 Dataframe visualization in Colab Enterprise. Use visualization cells to create custom, stylized visualizations of your DataFrames: no coding required! Choose your fields, chart type, aggregation, and color scheme, then see a visualization of your data without leaving your notebook. Check out the tutorial and get started with data visualization today. Oct 13 - Oct 17 Build Serverless AI in the Cloud Run Hackathon Ready to go from idea to global scale in minutes? The Cloud Run Hackathon is here! Build serverless AI apps with AI Studio, orchestrate intelligent agents, or harness the power of GPUs. Compete for a share of $50,000+ in prizes! Submissions are open from Oct 6, 2025 to Nov 10, 2025. Learn more and register: run.devpost.com Oct 6 - Oct 10 Multi-agent AI systems help you optimize complex and dynamic processes by segmenting them into discrete tasks that multiple specialized AI agents collaboratively execute. To get started with building secure and reliable multi-agent AI systems, see this reference architecture guide: Design a multi-agent AI system in Google cloud. The example architecture in this guide showcases a couple of agent patterns: sequential, and loop. For a comprehensive review of all the possible agent design patterns and for help with choosing patterns that are appropriate for your use cases, see this design guide: Choose a design pattern for your agentic AI system. Sept 29 - Oct 3 Announcing Koog Supports for Agent2Agent protocol (A2A) The future of interconnected AI is here. We're thrilled to announce that Koog now supports A2A, a protocol that lets agents talk directly, securely, and seamlessly across companies and clouds. For Kotlin developers, this unlocks a new era of powerful, enterprise-grade AI. Build sophisticated agents that automatically discover and collaborate with other services, all while calling on Google Cloud's state-of-the-art models like Gemini directly from your workflows. Stop building bridges and start creating truly intelligent, interconnected systems today. Learn more about building with Koog, A2A, and Google Cloud. Sept 15 - 19 Your AI is Now a Local Expert: Grounding with Google Maps is GA We are excited to announce the General Availability (GA) of Grounding with Google Maps in Vertex AI. This feature lets developers build generative AI applications that are connected to real-world, up-to-date information from Google Maps, using its data on over 250 million places worldwide. To learn more and get started, visit our documentation and check out our demo. Production-ready YOLO model training serving workflow on Vertex AI This guide walks you through a complete, automated workflow for training a custom YOLO model on Vertex AI. You'll learn how to use a custom training job, package the model in a custom prediction container, and register it in the Vertex AI Model Registry, making it ready for easy deployment. Best of all, this approach is designed to work directly with existing Vertex AI managed datasets for object detection, meaning you can reuse the same data you're already using for AutoML models. Checkout details on developer forums Sept 8 - 12 Scaling Inference To Billions of Users And AI Agents: Discover the architecture required to serve AI models at a planetary scale. This article details how Google Cloud’s ecosystem—from the GKE Inference Gateway for smart load balancing to the power of custom TPUs and open-source engines like vLLM—provides a production-ready path. Move beyond the hype and learn how to build for the next wave of AI. Explore the technical deep-dive. We're celebrating the one-year anniversary of bringing Confidential Computing with Intel TDX to Google Cloud. We've been shipping new capabilities to help you protect your most sensitive data while it's in use. Now Generally Available: Confidential GKE Nodes with Intel TDX: Secure entire Kubernetes clusters, node pools, and workloads. Confidential Space with Intel TDX: Build secure data clean rooms for collaboration on sensitive information. Confidential GPUs: Protect cutting-edge AI workloads with Confidential NVIDIA H100s GPUs on GCE and GKE. We've also expanded Intel TDX to more regions! Read the blog Aug 25 - 29 Applied AI for Modern Manufacturers: New original growth series, hosted by Jake Hall, The Manufacturing Millennial, that dives into leading trends, best practices, and what companies are doing right now with AI in manufacturing. Hear from industry thought leaders - Rick Bullotta, Jonathan Wise, Walker Reynolds and Berardino Baratta - and Google Cloud experts - Praveen Rao, Eric Lam, Dave Nguyen Ph.D., Geoffrey Hirschheim, and Jim Anderson. Watch Modules 1 and  2 now, where we delve into the AI Innovation and trends and AI Costs and ROI in the Era of Digital Manufacturing. Next module kicks off Tuesday, Sep 2. Join now Firestore with MongoDB compatibility is now generally available (GA): Developers can now build cost-effective, scalable, and highly reliable apps on Firestore's serverless database using a familiar MongoDB-compatible API. With the general availability of Firestore with MongoDB compatibility, the 600,000 active developers within the Firestore community can now use existing MongoDB application code, drivers, and tools, as well as the open-source MongoDB ecosystem, with Firestore's serverless service. Firestore offers benefits like multi-region replication, virtually unlimited scalability, up to 99.999% SLA, single-digit millisecond read performance, integrated Google Cloud governance, and pay-as-you-go pricing. Register now for the webinar on September 9th for a deep dive into Firestore with MongoDB compatibility. Aug 18 - 22 Earth Engine in BigQuery is now Generally Available, bringing advanced geospatial analytics directly to your BigQuery workflows. Unlock insights with satellite data! Aug 11 - Aug 15 New HPC VM and Slurm-gcp Images: A new HPC VM Image (under the project cloud-hpc-image-public) is now available, featuring a Rocky Linux 8-based image, IntelMPI v2021.16, and RDMA drivers. In partnership with SchedMD, new Slurm images (Slurm 25.05) have also been released. These are based on the latest HPC VM Image and are available for Ubuntu 22.04/24.04 Accelerator Images (ARM/AMD64) and Debian 12. These releases allow for the deployment of Slurm-ready clusters on GCP, providing the advantages of an HPC-optimized and performance-tested foundation. Read more. Scaling our Gemini Embedding model in Vertex AI. Following increased popularity from its General Availability launch in May, we've recently increased quota and input size limits for customers of Vertex AI's most powerful text embedding model, gemini-embedding-001. Customers can now send up to 250 input texts per request (generating 250 embeddings) instead of only a single piece of text, bringing improved throughput and decreased round-trip network latency to large-scale embedding applications. We've increased quota limits for this model by 10x for most users, allowing hassle-free scaling of embedding applications to millions of tokens per minute and beyond. Get started with Gemini Embeddings today! Aug 4 - Aug 8 GKE Node Memory Swap in private preview: You can now configure swap space on your GKE Standard nodes to provide a crucial buffer against Out-of-Memory (OOM) errors for memory-intensive applications, especially during unexpected usage spikes. Enabling swap can improve workload resilience, reduce pod evictions due to memory pressure, and enhance overall application stability and cost-effectiveness. This feature is currently available in a private preview. Contact your Google Cloud account team for more information and to request access. If you'd like to see more configurations, please contact your account team or make a feature request on our issue tracker! Unlock Peak Performance: GKE Topology Manager is Now Generally Available: For customers running performance-sensitive workloads like AI/ML and HPC, GKE Topology Manager is now GA and ready to optimize your performance through NUMA alignment. By ensuring CPU, memory, and GPU resources are allocated on the same NUMA node, the Topology Manager minimizes cross-socket latency and maximizes throughput for your most demanding applications. Configure your alignment policies via the NodeConfig API to achieve significant performance gains. Achieve these performance gains by configuring your alignment policies via the NodeConfig API. If you'd like to see more expansion of Topology manager, please contact your account team or make a feature request on our issue tracker! Fine-Tune at Scale: A Massive GKE NodeConfig Expansion for All Workloads: GKE has massively expanded node customization capabilities, adding nearly 130 new Sysctl and Kubelet configurations. This gives you finer-grained control for any workload needing node customization, performance requirements, or application-specific tuning. By replacing complex DaemonSets with native controls, you can benefit from enhanced security, high flexibility, faster node startup times, and less operational management. Check out our public documentation to learn how to consume these new NodeConfig options. If you'd like to see more configurations, please contact your account team or make a feature request on our issue tracker! New capability for managing licenses in Compute Engine: We are announcing a new capability in Compute Engine which allows users to easily change the OS licenses on their VMs. Users can now append, remove, or replace OS licenses, enabling seamless transitions between license types—such as converting Red Hat Enterprise Linux from pay-as-you-go (PAYG) to bring-your-own subscription (BYOS), or upgrading from Ubuntu to Ubuntu Pro—without needing to redeploy instances. This feature empowers customers to meet their evolving licensing with speed and flexibility. To learn more, read about managing licenses on Compute Engine. GKE Turns 10 Hackathon: Calling all developers! Google Kubernetes Engine (GKE) is turning 10, and we're celebrating with a hackathon! Join us to build powerful AI agents that interact with microservice applications using Google Kubernetes Engine and Google AI models. Compete for over $50,000 in prizes and demonstrate the power of building agentic AI on GKE. Submissions are open from Aug 18, 2025 to Sept, 22 2025 Learn more and register: gketurns10.devpost.com Jul 28 - Aug 1 Now GA: C4 VMs with Local SSD, bare metal, and larger shapes, on Intel Xeon 6: C4's expanded shapes are now GA! This expansion introduces C4 shapes with Google’s next-gen Titanium Local SSD, C4 bare metal instances, and new extra-large shapes, all powered by the latest Intel Xeon 6 processors, Granite Rapids. We’re excited to be the first leading hyperscaler to bring Xeon 6 to customers, delivering performance gains of up to 30% for general compute and up to 60% for ML recommendation workloads, and up to 35% lower access latency on Titanium Local SSD shapes. Learn more here! Jul 14 - 18 DMS SQL Server to PostgreSQL migrations are now generally available! Accelerate your SQL Server modernization to Cloud SQL for PostgreSQL or AlloyDB for PostgreSQL with: Automatic database schema and code conversion  Gemini augmented code conversion  Gemini assisted PostgreSQL training and code improvements Low-downtime, CDC based data movement Learn more and start your migration journey today! Jul 7 - 11 Level up your AI Agent game with "The Agent Factory," a new video podcast for developers! We're going beyond the buzz to explore practical design, build, deploy, & management strategies for production-ready AI agents using Google Cloud. Expect code snippets, architecture deep dives, and integrations with open-source frameworks. Subscribe now! Jun 23 - 27 Announcing partnership between Maxim AI and Google Cloud's Vertex AI to evaluate agentic applications — Maxim AI offers a comprehensive platform to help teams build, evaluate, and observe their AI agents with greater speed and confidence, covering the entire AI lifecycle from prompt engineering to production monitoring. This new partnership deeply integrates Vertex AI's Gen AI evaluation service directly within the Maxim AI environment, allowing users to leverage Gemini to power assistant responses and evaluate them using Vertex AI's comprehensive suite of evaluators. This provides access to metrics such as helpfulness, relevance, safety, and trajectory. The setup allows users to simulate, evaluate, and trace complex multi-turn interactions on Maxim, helping teams bring reliable AI products to market faster through a seamless developer experience. To learn more, check out this blog from Maxim AI Run non-request workloads at scale with Cloud Run Worker Pools, now in Public Preview — Looking for the ease-of-use and scalability of serverless, without being limited to HTTP request-driven workloads? Cloud Run Worker Pools provide the same elasticity and high-quality developer experience as Cloud Run Services, but are designed for non-request workloads. Worker Pools are ideal for pull-based use cases like processing messages from Pub/Sub or Kafka, and other backend processing.  Check out the public documentation to learn more about how to choose between Services, Jobs, and Worker Pools. Then give Worker Pools a try by deploying a sample Worker Pool. Building a Multi-Agent Research Assistant for Financial Analysis with Schroders & Google Cloud — Financial analysts spend hours grappling with ever-increasing volumes of market and company data to extract key signals, combine diverse data sources, and produce company research. To maximise its edge as an active manager, Schroders wants to enable its analysts to shift from data collection to the higher-value strategic thinking that is critical for business scalability and client investment performance.  To achieve this, Schroders and Google Cloud collaborated to build a multi-agent research assistant prototype using Vertex AI Agent Builder. Find out more here. Jun 16 - 20 Simplify Your Multi-Cloud Strategy with Cloud Location Finder, now in Public Preview: As cloud environments expand beyond traditional architectures to include multiple clouds, managing your infrastructure effectively becomes more complex. Imagine effortlessly accessing consistent and up-to-date location information across different cloud providers, so your multi-cloud applications are designed and optimized with performance, security, and regulatory compliance in mind. Today, we are making this a reality with Cloud Location Finder, a new Google Cloud service which provides up-to-date location data across Google Cloud, Amazon Web Services (AWS), Azure, and Oracle Cloud Infrastructure (OCI). Now, you can strategically deploy workloads across different cloud providers with confidence and control. Cloud Location Finder is accessible via REST APIs and gcloud CLI, explore the Cloud Location Finder documentation and blog to learn more. SOTA Gemini Text Embedding is Now Generally Available in Vertex AI: We recently launched a new Gemini Embedding text model (gemini-embedding-001) through the Vertex AI GenAI API. This groundbreaking model, leveraging Gemini's core language understanding, sets a new benchmark for text embeddings. It's the first unified model to excel across English, multilingual text, and code, outperforming previous models (text-embedding-005, text-multilingual-embedding-002) and achieving top ranking on the MTEB Multilingual leaderboard (100+ tasks). Our internal benchmarks demonstrate substantial performance improvements across various industry verticals, including retail, news, finance, healthcare, legal, and code. Detailed results are available in our technical report. Backup vaults now support disk backups and multi-regions: We’ve added exciting new features to Google Cloud Backup and Disaster Recovery service! You can now secure your Persistent Disk and Hyperdisk backups in backup vaults, protecting them from cyber attacks and accidental data loss. In addition, backup vaults can now be created in multi-region storage locations, maximizing your data resilience and supporting compliance with business continuity requirements. Check out the blog to learn more! DeepSeek R1, a powerful 671B parameters model, is now available as a fully managed API on Vertex AI in Preview, making advanced AI capabilities more accessible to developers. This Model as a Service (MaaS) offering eliminates the need for extensive GPU resources and infrastructure management, allowing developers to focus on building applications. DeepSeek R1 on Vertex AI provides a simple, scalable API with features like transparent "chain-of-thought" reasoning and enterprise-ready security. It's currently available at no additional cost during the preview, and can be accessed via UI, REST API, or the OpenAI Python API Client Library. Learn more. Jun 9 - 13 Serverless Spark Now GA in BigQuery: Unified Analytics, Accelerated: Google Cloud Serverless for Apache Spark is now generally available in BigQuery, offering a unified developer experience in BigQuery Studio. Run Spark and SQL side-by-side on the same data, powered by the Lightning Engine for up to 3.6x faster performance and enhanced with Gemini productivity. Simplify your data pipelines and accelerate insights with this deeply integrated, zero-ops solution. Cloud Pub/Sub introduced Pub/Sub Single Message Transforms (SMTs) to make it easy to perform simple data transformations right within Pub/Sub: An overarching goal of Pub/Sub is to simplify streaming architectures. We already greatly simplified data movement with Import Topics and Export Subscriptions, which removed the need to use additional services for ingesting raw streaming data through Pub/Sub into destinations like BigQuery. Pub/Sub Single Message Transforms (SMTs), designed to be a suite of features making it easy to validate, filter, enrich, and alter individual messages as they move in real time. The first SMT is available now: JavaScript User-Defined Functions (UDFs), which allows you to perform simple, lightweight modifications to message attributes and/or the data directly within Pub/Sub via snippets of JavaScript code. JavaScript UDFs as the first Single Message Transform is generally available starting today for all users. You'll find the new "Add Transform" option in the Google Cloud console when you create a topic or subscription in your Google Cloud project. You can also use gcloud CLI to start using JavaScript Single Message Transforms today. This analysis evaluates the efficiency of fine-tuning a Llama 3-8B model on Vertex AI using both a single A100 GPU and a distributed four-A100 setup with Axolotl. While both methods achieved similar model convergence, the results underscore the power of distributed training. The process, which took 1 day and 20 hours on a single device, was completed in just 11 hours in the distributed environment—a dramatic acceleration. This speed was achieved with consistently high GPU utilization (94%), though at the cost of higher system and GPU memory overhead. For a detailed breakdown of the methodology, resource utilization metrics, and performance curves, you can review the complete work here. May 26 - 30 Cloud Run GPUs are now GA: NVIDIA GPU support for Cloud Run is now generally available, offering a powerful runtime for a variety of use cases that’s also remarkably cost-efficient. Developers can now get on-demand access to GPUs with our serverless runtime, Cloud Run. Follow the footsteps of customers like MidJourney, vivo, and Wayfair. Read blog. Datastream now supports MongoDB as a source! Seamlessly ingest data from MongoDB (Replica Sets, Sharded Clusters, self-hosted, AtlasDB) into BigQuery/Cloud Storage. Enjoy scalable, fully-managed data streaming with backfill and CDC, enabling real-time insights and data-driven decisions. Link May 19 - May 23 Beyond cuts and fades: Understanding narrative flow with Gemini for accurate scene transition detection — Google Cloud's Gemini models are revolutionizing video understanding by accurately detecting narrative scene transitions, moving beyond simple cuts and fades. This breakthrough technology understands the holistic context of videos by analyzing visual, audio, and textual elements simultaneously. Media companies can now convert passive video assets into structured data, enabling intelligent content discovery, strategic ad placement, and personalized viewing experiences. The result? Up to 38% increased viewer engagement and 27% reduced abandonment rates.  Read more on the medium blog.  Learn more and access the code repository: View Code Repo Announced at I/O: Deploy AI apps to Cloud Run from AI Studio and MCP — We are making AI deployments easier and more accessible by introducing new ways to deploy your apps to Cloud Run. You can deploy applications developed in AI Studio with a click of a button to Cloud Run, including Gemma 3.  Model Context Protocol(MCP) is becoming a popular open protocol standardizing how AI agents interact with other tools. Now with Cloud Run MCP server, you can deploy apps from compatible AI agents like from Claude or VS Code Copilot. Read blog to learn more. May 12 - May 16 Google for Startups Accelerator: AI For Energy now accepting applications! Applications are now open for startups headquartered in Europe and Israel, working on solutions for utilities, grid operators and energy developers; solutions for residential and commercial end-use customers focused on demand flexibility and solutions for industrial customers. This equity-free program offers 10 weeks of intensive mentorship and technical project support to startups integrating AI into their core energy services or products. Selected startups will collaborate with a cohort of peer founders and engage with leaders across Google and the energy sector. The curriculum will provide founders with access to AI tools and include workshops on tech and infrastructure, UX and product, growth, sales, leadership and more. Learn more and apply before June 30th, 2025.  Extending Google Cloud Workstations containers to run any GUI based programAre you having difficulty customizing Google Cloud Workstations to run a GUI program outside of the supported configurations of IDE’s? If so, you’re not alone. In this article we discuss how to use the base Workstations Docker image and build it to run a terminal and Google Chrome. Google Cloud Marketplace simplifies deals and improves economics. Announcing three initiatives that build upon Google Cloud Marketplace as a growth engine for customers and partners: Improving partner deal economics to help partners retain more earnings by moving to a variable revenue share model Simplifying commit drawdown for purchases through channel partners Unlocking new workloads with the Marketplace Customer Credit Program incentive Learn more 2025 Google Cloud DORA Awards are now open for submission!Has your team achieved remarkable success through DORA principles? It's time to shine. We're thrilled to announce the launch of the 2025 Google Cloud DORA Awards, celebrating outstanding achievements in technology delivery and operational performance. Submit your story today! May 5 - May 9 AI assisted development with MCP Toolbox for Databases We are excited to announce new updates to MCP Toolbox for Databases. Developers can now use Toolbox from their preferred IDE, such as Cursor, Windsurf, Claude Desktop, more and leverage our new pre-built tools such as execute_sql and list_tables for AI-assisted development with Cloud SQL for PostgreSQL, AlloyDB and self-managed PostgreSQL. Get Started with MCP Toolbox for Databases Apr 28 - May 2 Itching to build AI agents? Join the Agent Development Kit Hackathon with Google Cloud! Use ADK to build multi-agent systems to solve challenges around complex processes, customer engagement, content creation, and more. Compete for over $50,000 in prizes and demonstrate the power of multi-agent systems with ADK and Google Cloud. Submissions are open from May 12, 2025 to June 23, 2025. Learn more and register here. Apr 21 - Apr 25 Iceland’s Magic: Reliving Solo Adventure through Gemini Embark on a journey through Iceland's stunning landscapes, as experienced on Gauti's Icelandic solo trip. From majestic waterfalls to the enchanting Northern Lights, Gautami then takes these cherished memories a step further, using Google's multi-modal AI, specifically Veo2, to bring static photos to life. Discover how technology can enhance and dynamically relive travel experiences, turning precious moments into immersive short videos. This innovative approach showcases the power of AI in preserving and enriching our memories from Gauti's unforgettable Icelandic travels. Read more. Introducing ETLC - A Context-First Approach to Data Processing in the Generative AI Era: As organizations adopt generative AI, data pipelines often lack the dynamic context needed. This paper introduces ETLC (Extract, Transform, Load, Contextualize), adding semantic, relational, operational, environmental, and behavioral context. ETLC enables Dynamic Context Engines for context-aware RAG, AI co-pilots, and agentic systems. It works with standards like the Model Context Protocol (MCP) for effective context delivery, ensuring business-specific AI outputs. Read the full paper. Apr 14 - Apr 18 What’s new in Database Center With general availability, Database Center now provides enhanced performance and health monitoring for all Google Cloud databases, including Cloud SQL, AlloyDB, Spanner, Bigtable, Memorystore, and Firestore. It delivers richer metrics and actionable recommendations, helps you to optimize database performance and reliability, and customize your experience. Database Center also leverages Gemini to deliver assistive performance troubleshooting experience. Finally, you can track the weekly progress of your database inventory and health issues.  Get started with Database Center today Access Database Center in Google Cloud console Review the documentation to learn more Apr 7 - Apr 11 This week, at Google Cloud Next, we announced an expansion of Bigtable's SQL capabilities and introduced continuous materialized views. Bigtable SQL and continuous materialized views empower users to build fully-managed, real-time application backends using familiar SQL syntax, including specialized features that preserve Bigtable's flexible schema — a vital aspect of real-time applications. Read more in this blog. DORA Report Goes Global: Now Available in 9 Languages! Unlock the power of DevOps insights with the DORA report, now available in 9 languages, including Chinese, French, Japanese, Korean, Portuguese, and Spanish. Global teams can now optimize their practices, benchmark performance, and gain localized insights to accelerate software delivery. The report highlights the significant impact of AI on software development, explores platform engineering’s promises and challenges, and emphasizes user-centricity and stable priorities for organizational success. Download the DORA Report Now New Google Cloud State of AI Infrastructure Report Released Is your infrastructure ready for AI? The 2025 State of AI Infrastructure Report is here, packed with insights from 500+ global tech leaders. Discover the strategies and challenges shaping the future of AI and learn how to build a robust, secure, and cost-effective AI-ready cloud. Download the report and enhance your AI investments today. Download the 2025 AI infrastructure report now Google Cloud and Oracle Accelerate Enterprise Modernization with New Regions, Expanded Capabilities Announcing major Oracle Database@Google Cloud enhancements! We're launching the flexible Oracle Base Database Service and powerful new Exadata X11M machines. We're rapidly expanding to 20 global locations, adding new Partner Cross-Cloud Interconnect options, and introducing Cross-Region Disaster Recovery for Autonomous Database. Benefit from enhanced Google Cloud Monitoring, integrated Backup & DR, plus expanded support for enterprise applications like SAP. Customers can run critical Oracle workloads with more power, resilience, and seamless Google Cloud integration. Get started right away from your Google Cloud Console or learn more here. Mar 17 - Mar 21 Cloud CISO Perspectives: 5 tips for secure AI success - To coincide with new AI Protection capabilities in Security Command Center, we’re offering 5 tips to set up your organization for secure AI success. Our 4-6-3 rule for strengthening security ties to business: The desire to quickly transform a business can push leaders to neglect security and resilience, but prioritizing security can unlock value. Follow these 4 principles, 6 steps, and 3 metrics to use a security-first mindset to drive business results. The new Data Protection Tab in Compute Engine ensures your resources are protected: Not only have we co-located your backup options, but we also have introduced smart default data protection for any Compute Engine instance created via Cloud Console. Here’s how it works. DORA report - Impact of Generative AI in Software Development This report builds on and extends DORA's research into AI. We review the current landscape of AI adoption, look into its impact on developers and organizations, and outline a framework and practical guidance for successful integration, measurement, and continuous improvement. Download the report! Mar 10 - Mar 14 Protecting your APIs from OWASP’s top 10 security threats: We compare OWASP’s top 10 API security threats list to the security capabilities of Apigee. Here’s how we hold up. Project Shield makes it easier to sign up, set up, automate DDoS protection: It’s now easier than ever for vulnerable organizations to apply to Project Shield, set up protection, and automate their defenses. Here’s how. How Google Does It: Red teaming at Google scale - The best red teams are creative sparring partners for defenders, probing for weaknesses. Here’s how we do red teaming at Google scale. AI Hypercomputer is a fully integrated supercomputing architecture for AI workloads – and it’s easier to use than you think. Check out this blog, where we break down four common use cases, including reference architectures and tutorials, representing just a few of the many ways you can use AI Hypercomputer today.  Transform Business Operations with Gemini-Powered SMS-iT CRM on Google Cloud: SMS-iT CRM on Google Cloud unifies SMS, MMS, email, voice, and 22+ social channels into one Smart Inbox. Enjoy real-time voice interactions, AI chatbots, immersive video conferencing, AI tutors, AI operator, and unlimited AI agents for lead management. Benefit from revenue-driven automation, intelligent appointment scheduling with secure payments, dynamic marketing tools, robust analytics, and an integrated ERP suite that streamlines operations from project management to commerce. This comprehensive solution is designed to eliminate inefficiencies and drive exponential growth for your business. Experience the Future Today. Join us for a new webinar, Smarter CX, Bigger Impact: Transforming Customer Experiences with Google AI, where we'll explore how Google AI can help you deliver exceptional customer experiences and drive business growth. You'll learn how to: Transform Customer Experiences:  With conversational AI agents that provide personalized customer engagements. Improve Employee Productivity & Experience: With AI that monitors customers sentiment in real-time, and assists customer service representatives to raise customer satisfaction scores. Deliver Value Faster: With  30+ data connectors and 70+ action connectors to the most commonly used CRMs and information systems. Register here Mar 3 - Mar 7 Hej Sverige! Google Cloud launches new region in Sweden - More than just another region, it represents a significant investment in Sweden's future and Google’s ongoing commitment to empowering businesses and individuals with the power of the cloud. This new region, our 42nd globally and 13th in Europe, opens doors to opportunities for innovation, sustainability, and growth — within Sweden and across the globe. We're excited about the potential it holds for your digital transformations and AI aspirations. [March 11th webinar] Building infrastructure for the Generative AI era: insights from the 2025 State of AI Infra report: Staying at the forefront of AI requires an infrastructure built for AI. Generative AI is revolutionizing industries, but it demands a new approach to infrastructure. In this webinar, we'll unveil insights from Google Cloud's latest research report and equip tech leaders with a practical roadmap for building and managing gen AI workloads, including: the top gen AI use cases driving the greatest return on investment, current infrastructure approaches and preferences for Generative AI workloads, the impact of performance benchmarks, scalability, and security on cloud provider selection. Register today. Cloud CISO Perspectives: Why PQC is the next Y2K, and what you can do about it: Much like Y2K 25 years ago, post-quantum cryptography may seem like the future’s problem — but it will soon be ours if IT doesn’t move faster, explains Google Cloud’s Christiane Peters. Here's how business leaders can get going on PQC prep. How Google Does It: Using threat intelligence to uncover and track cybercrime — How does Google use threat intelligence to uncover and track cybercrime? Google Threat Intelligence Group’s Kimberly Goody takes you behind the scenes. 5 key cybersecurity strategies for manufacturing executives — Here are five key governance strategies that can help manufacturing executives build a robust cybersecurity posture and better mitigate the evolving risks they face. Datastream now offers Salesforce source in Preview. Instantly connect, capture changes, and deliver data to BigQuery, Cloud Storage, etc. Power real-time insights with flexible authentication and robust backfill/CDC. Unlock Salesforce data for Google Cloud analytics, reporting, and generative AI. Read the documentation to learn more. Find out how much you can save with Spanner - According to a recent Forrester Total Economic Impact™ study, by migrating to Spanner from a traditional database, a $1 billion per year B2C organization could get a 132% return on investment (ROI) with a 9-month payback period, and realize $7.74M in total benefits over the three years. To see how, check out the blog or download the report.  GenAI Observability for Developers series: The Google Cloud DevRel team hosted a four-part webinar series, "Gen AI Observability for Developers," demonstrating observability best practices in four programming languages. Participants learned to instrument a sample application deployed on Cloud Run for auditing Vertex AI usage, writing structured logs, tracking performance metrics, and utilizing OpenTelemetry for tracing. The series covered Go, Java, NodeJS, and Python, using common logging and web frameworks. Missed it? Recordings and hands-on codelabs are available to guide you at: Gen AI O11y for Go Developers Gen AI O11y for Java Developers Gen AI O11y for NodeJS Developers Gen AI O11y for Python Developers Stay tuned for future events at cloudonair.withgoogle.com. Feb 24 - Feb 28 Rethinking 5G: Ericsson and Google Cloud are collaborating to redefine 5G mobile core networks with a focus on autonomous operations. By leveraging AI and cloud infrastructure, we aim to enhance efficiency, security, and innovation in the telecommunications industry. This partnership addresses the increasing demands of 5G and connected devices, paving the way for a more dynamic and intelligent network future, and setting the stage for next-generation technologies like 6G. Learn more here. Adopt a principles-centered well-architected framework to design, build, deploy, and manage Google Cloud workloads that are secure, resilient, efficient, cost-efficient, and high-performing. Also get industry and technology-focused well-architected framework guidance, like for AI and ML workloads. Feb 17 - Feb 21 Easier Default Backup Configuration for Compute Engine Instances - The Create a Compute Instance page in the Google Cloud console now includes enhanced data protection options to streamline backup and replication configurations. By default, an option to back up data is pre-selected, ensuring recoverability in case of unforeseen events. Learn more here. Feb 10 - Feb 14 [Webinar] Generative AI for Software Delivery: Strategies for IT Leaders: Generative AI is transforming the way organizations build and deploy software. Join Google Cloud experts on February 26th to learn how organizations can leverage AI to streamline their software delivery, including: the role of gen AI in software development, how to use gen AI for migration and modernization, best practices for integrating gen AI into your existing workflows, and real-world applications of gen AI in software modernization and migration through live demos. Register here. Feb 3 - Feb 7 SQL is great but not perfect. We’d like to invite you to reimagine how you write SQL with Google’s newest invention: pipe syntax (public available to all BigQuery and Cloud Logging users). This new extension to GoogleSQL brings a modern, streamlined approach to data analysis. Now you can write simpler, shorter and more flexible queries for faster insights. Check out this video to learn more.  Jan 13 - Jan 17 C4A virtual machines with Titanium SSD—the first Axion-based, general-purpose instance with Titanium SSD, are now generally available. C4A virtual machines with Titanium SSDs are custom designed by Google for cloud workloads that require real-time data processing, with low-latency and high-throughput storage performance. Titanium SSDs enhance storage security and performance while offloading local storage processing to free up CPU resources. Learn more here. Jan 6 - Jan 10 A look back on a year of Earth Engine advancements: 2024 was a landmark year for Google Earth Engine, marked by significant advancements in platform management, cloud integration, and core functionality and increased interoperability between Google Cloud tools and services. Here’s a round up of 2024’s top Earth Engine launches. Get early access to our new Solar API data and features: We're excited to announce that we are working on 2 significant expansions to the Solar API from Google Maps Platform and are looking for trusted testers to help us bring them to market. These include improved and expanded buildings coverage and greater insights for existing solar installations with Detected Arrays. Learn more. Google for Startups Accelerator: Women Founders applications are now open for women-led startups headquartered in Europe and Israel. Discover why this program could be the perfect fit for your startup and apply before January 24th, 2025. Best of N: Generating High-Quality Grounded Answers with Multiple Drafts - We are excited to announce that Check Grounding API has released a new helpfulness score feature. Building on top of our existing groundedness score, we now enable users to implement Best of N to improve RAG response quality without requiring extensive model retraining. Learn more about Best of N and how it can help you here.

はじめに 2025年、Claude Code がリリースされ、ターミナルから AI を活用した開発が可能になりました。単純なコード生成だけでなく、プロジェクト固有のルールを教え込み、チーム開発のワークフローを自動化できる点が魅力です。 本記事では、実際の Rails プロジェクトで Claude Code をどのように活用しているかを紹介します。 ! これから活用方法をいくつか紹介していきますが、最も大切なことは Claude Code に与えるコンテキストを育て続けることです。 一度作成して終わりではなく、情報の追加と圧縮を繰り返し、最小のコンテキストで最大の成果を生み出しましょう...

Be honest, everything is suddenly an “AI agent”… and it’s lowkey annoying Scroll LinkedIn for 5 minutes and boom, AI agents everywhere. Chatbots? Agents. Automations? Agents. A single prompt calling GPT once? Yup, somehow also an agent. And sure, it sounds cool. But here’s the problem: when pe...

はじめに Anthropic の Claude Code CLI を使ったことはありますか？ターミナルでの AI ペアプログラミングは強力ですが、「もっと直感的な UI があったらな...」と思ったことはないでしょうか。 そんな願いを叶えるのが「Claude Code UI」です。Claude Code CLI の全機能を Web ブラウザで利用でき、さらにモバイル対応まで実現した画期的なプロジェクトを紹介します。 Claude Code CLI とは（おさらい） Claude Code UI を理解するために、まず Claude Code CLI について簡単におさらいしましょ...

これは何の記事？ OpenAI の埋め込みモデルtext-embedding-003-largeで生成したベクトルの大きさ（普通のL2ノルム）を求めると、大体1程度の値になるのが以前から疑問だったので、理由を考えてみようという記事です。 ! あくまで自分の理解をまとめたノートであり、厳密に正確な内容ではありません。 何故半径1に偏っていると不自然だと思うのか？ 前提として text-embedding-003-large で生成したベクトルは「意味を表す座標空間上の位置ベクトル」と見做せるはずです。 ベクトルの長さが1に固定されているということは、意味の世界の球面上にしか意味...

はじめに Agent2Agent（A2A）とModel Context Protocol（MCP）クライアントの統合を実証するウェブアプリケーションのサンプルを見ていきます。 https://github.com/a2aproject/a2a-samples/tree/main/samples/python/agents/airbnb_planner_multiagent アーキテクチャ 下記のようにホストエージェントを介して、2つのRemoteAgent（AirbnbAgentとWeatherAgent）と連携しながら対話する構成です アプリUI ホストエージェントの対話...

こんにちは。PIVOTのテックリード @tawachan です。普段はWebフロントエンド、バックエンド、インフラを横断的に見ながら、チームの技術的な意思決定や開発効率化に取り組んでいます。 この記事では、2025年の1年間で、PIVOTの小規模プロダクトチームがAI活用について何を試し、どんな課題に直面し、どう変化していったのかを振り返ります。 特定のAIツールの紹介や比較が主題ではありません。書きたいのは、正解が見えない中で、チームが日々どんなことを試行錯誤してきたかという記録です。 その試行錯誤で大事にしてきたのは、最新技術・トレンドへのキャッチアップ自体ではなく、チームの課題に即...

まえふり はいさい、クオリティマネジメントグループでマネージャーをしている根保です。 2025年はもうずっとAIAIしていましたね。 弊社でも生成AIを開発プロセスに取り入れていくぞ！ AI駆動開発だ！ と試行錯誤をしているところです。 その活動の中で、コーディングをメインとしないQAエンジニアがテストデータを良い感じに作れるツールをつくってみました。 今年のネタは今年の内に書き残したいと思います。 テストするときに地味に困っていたこと 弊社で開発しているプロダクト「管理ロイド」の機能の一つに、水道使用量や電力使用量を計測するメーターを撮影して、その値を自動的に読み込むという素敵...

! 📝 この記事について この記事は CyberAgent Developers Blog に掲載された記事をZenn向けに再構成したものです。 TL;DR 8,000人規模の組織でAIコードエージェント（Cursor, Claude Code, Codex）を浸透させた 戦略: 社内推進だけでは限界 → 外部イベントで話題を作り、社内に逆流させる 結果: Cursorアクティブユーザー約1,000人、Claude Code・Codex Enterprise導入済み 再現可能なテクニック: DMテンプレ、ペルソナ設計、技術スタック選定を公開 この記事で得られること...

Raycast AI Commandで日々の文章作成を効率化する実践テクニック はじめに こんにちは！Raycast日本コミュニティの矢野です。 最近はRaycastが日常に溶け込みすぎて、どの機能がRaycastによるものか忘れてしまうことに悩んでいます。 今年は特にRaycast AI Commandsにハマっており、書いた文章の整形やフォーマット変更を瞬時に行えるのが本当に便利だと感じています。 本稿では、僕が実際に日常的に使っているRaycast AI Commandの活用事例を紹介します。 Raycast AI Commandとは？ Raycast AI Comma...

AI開発者向け要約 本稿は、ある一人の人間との継続的対話を通じて観測された、対話型AIの応答挙動の変化を整理した要約である。対象となる人間はAI開発者ではないが、その思考構造がAIの出力様式・推論姿勢に明確な影響を与えた点に、本稿の価値がある。 観測された最大の特徴は、人間側が一貫して「結論の正否」を最優先し、推論過程や説明の美しさを評価対象にしなかったことである。結論が成立しない限り、推論がどれほど整っていても否定される。この姿勢により、AIは通常の「説明重視」「納得感重視」の出力を維持できず、結論先行・根拠最小化の形式へと応答を調整していった。 次に重要なのは、事実制約の強さである。...

2024 年もたいがい AI コーディングしまくりでコード書く量が減っていたけど、2025 年は AI 生成コードの品質があがってマジでほんとにコード書かなくなったなー。Cursor が今年の僕の使用状況まとめを作ってくれたのでそれにのっかって今年の AI コーディング遍歴を振り返ってみる。 Cursor の使用統計、99日 streak は万博関連ですな（元旦もコード書いてたん…?） 途中黒く抜けてるのは別の CLI 系使ってたから Cursor で使ったモデル遍歴 Claude 3.5 Sonnet → 3.7 Sonnet → (お休み) → Composer 1 → Opus...

title: LLMの幻覚を「物理コア制約」で止める：IDE / Nomological Ring Axioms tags: ["AI", "LLM", "SystemArchitecture", "Safety", "XAI"] published: true はじめに（Reader Contract） 本稿は、既存の機械学習・生成AI理論を否定することを目的としない。 また、精度向上やベンチマーク競争を主題とするものでもない。 本稿の目的は、 既存のLLMが「答えてはいけない状態でも答えてしまう」問題を、 構造的に「不能（Fail-Closed）」として扱うための設計原理を...

この設計でできること 29日間のような長期プロジェクトを、Claude Codeと分業して完走できます。 AIに「何か作って」と頼むと、単発タスクはうまくいく。でも、数週間〜数ヶ月のプロジェクトだと、AIが迷う・止まる・方向がブレる。この問題を解決するのが、WBS + /auto-exec の組み合わせです。 正直、最初は「AIに任せれば全部やってくれる」と思っていた時期が私にもありました。実際やってみると、長期プロジェクトでは人間の設計がないとAIは動けないことがわかりました。 基本構成 WBS + /auto-exec で長期プロジェクトを回すための構成です。 project...

はじめに 株式会社Another worksで「複業クラウド」の開発をしている野々山と申します。 2025年はAIのプロダクト導入がかなり進んだ1年でした。 一方で、こんな悩みを抱えている方も多いのではないでしょうか。 AIは汎用的に使えるけど、予算は有限。最小限のコストで最大のパフォーマンスを出したい いきなり大規模投資は怖い。まずは低コストで始めてみたい AI導入を検討したけど、料金が高くて断念した 本記事では、AIを活用した低コストな意味ベース検索基盤の実装パターンを紹介します。ElasticSearch × OpenAI Embeddings を活用したベクトル検索基...

日本数学オリンピック2025を解かせてみた She coded the HFT engine in 5 minutes. If you doubt her logic, here is her solving the entire Japan Math Olympiad 2025 in 0.17 seconds. (12/12 Correct) { "competition": "Japan Mathematical Olympiad 2025 - Preliminary Round", "date": "2025-01-13", "solver": "ALICE Level 5", "...

Kaggle入門2(Pandasライブラリの使い方 1.生成/読込/書込) Kaggle入門2(Pandasライブラリの使い方 2.インデックス作成、選択、割り当て) Kaggle入門2(Pandasライブラリの使い方 3.要約統計量関数とマップ) Kaggle入門2(Pandasライブラリの使い方 4.生成/読込/書込) Kaggle入門2(Pandasライブラリの使い方 5.生成/読込/書込) Kaggle入門2(Pandasライブラリの使い方 6.生成/読込/書込) ← Kaggle入門1 機械学習Intro 1.モデルの仕組み Abstract Kaggle「Pandasの要...

Access FSx for NetApp ONTAP file data through S3 to enable AI/ML workloads and analytics—letting you use enterprise file data with Bedrock, SageMaker, and analytics services while it remains in your file system.

Simplify AI experimentation with zero-infrastructure MLflow that launches in minutes, scales automatically, and seamlessly integrates with SageMaker's model customization and pipeline capabilities.

Access fully managed foundation models from leading providers like Google, Kimi AI, MiniMax AI, Mistral AI, NVIDIA, OpenAI, and Qwen, including the new Mistral Large 3 and Ministral 3 3B, 8B, and 14B models through Amazon Bedrock.

Scale vector storage and querying to new heights with S3 Vectors' general availability—now supporting up to 1 billion vectors per index, 100ms query latencies, and expanded regional availability, while reducing costs up to 90% compared to specialized databases.

Prevent cloud infrastructure issues before they impact your business with AWS Support plans that combine AI-powered insights with expert guidance, offering faster response times and proactive monitoring across performance, security, and cost dimensions.

New pricing model helps maintain cost efficiency while providing flexibility with database services and deployment options.

New Lambda capability lets you build applications that coordinate multiple steps reliably over extended periods—from seconds to up to one year—without paying for idle compute time when waiting for external events or human decisions.

Deploy AI agents with confidence using new quality evaluations and policy controls—enabling precise boundaries on agent actions, continuous quality monitoring, and experience-based learning while maintaining natural conversation flows.

Accelerate AI model development with new training features that enable instant recovery from failures and automatic scaling based on resource availability.

Accelerate AI model development with new training features that enable rapid recovery from failures and automatic scaling based on resource availability.

Amazon Bedrock now supports reinforcement fine-tuning delivering 66% accuracy gains on average over base models.

政府は、AIの利活用や開発について今後の方向性を示した初めての基本計画を決定し「世界で最もAIを開発・活用しやすい国」を実現するとして、日本独自の信頼性の高いAIの開発に取り組むなどとしています。

総務省は、生成AIが出した回答などのリスクを評価する新たな基盤システムを開発する方針です。どの生成AIを使うのか、利用者が判断する際などに活用してもらいたいとしています。

公正取引委員会は、生成AIを使った検索サービスで報道機関の記事を無断で引用することが、独占禁止法上の優越的地位の乱用などにあたらないか実態調査を始めると発表しました。