Google turns Chrome into an agent platform

Google is trying to turn the enterprise browser into a place where artificial intelligence does work, not just displays it. In Chrome Enterprise, Gemini is moving from a sidebar helper toward an agent that can act across web tasks. (cloud.google.com 1) (cloud.google.com 2) At Google Cloud Next ’26 on April 22, Google said Auto Browse in Gemini in Chrome is available for eligible Google Workspace users in the United States. Google’s release notes say the feature starts rolling out in Chrome 143 on macOS and Windows, with an enterprise policy called GeminiActOnWebSettings available at launch. (chromeenterprise.google) (cloud.google.com) Google has been describing the browser as “intelligent, agentic and more secure,” with Gemini in Chrome already able to summarize reports, extract information from videos and work across tabs. Release notes also show multi-tab context controls and later access to Google Drive files as shareable context for AI features. (cloud.google.com) (chromeenterprise.google) An agent, in Google’s framing, is software that can take a goal and handle the steps in between. The browser matters because employees already spend much of their day inside web apps, making Chrome a natural control point for both automation and security. (cloud.google.com) (blog.google) Google paired that browser push with a bigger platform change. On April 22, it introduced Gemini Enterprise Agent Platform as the successor to Vertex AI for building, scaling, governing and optimizing agents, combining model services with orchestration, integration, security and DevOps tools. (cloud.google.com) (blog.google) Google also said the Gemini Enterprise app now sits on that same platform, so companies can discover, create, share and run agents in one environment. The company said the system includes connectors to enterprise data, native permission controls, and observability and traceability features for audits. (cloud.google.com 1) (cloud.google.com 2) That is the practical shift in this week’s news: Chrome is becoming one front end for agentic work, while Agent Platform is the back end that decides what those agents can reach, log and execute. Google is selling the pair as a governed system rather than a loose collection of copilots. (cloud.google.com 1) (cloud.google.com 2) Google is also trying to show this is not just a product diagram. Mars said on April 22 that it is expanding its partnership with Google Cloud and designating Gemini Enterprise as the primary AI operating system for its global workforce, with rollout continuing through 2026. (mars.com) (prnewswire.com) Chrome Enterprise’s own product pages now describe the browser as a “secure AI browser,” and Google prices Chrome Enterprise Premium at $6 per user monthly for advanced controls. That pricing and policy layer signals where Google expects companies to make the trade: more automation inside the browser, in exchange for tighter admin control over data and actions. (chromeenterprise.google) (chromeenterprise.google) The open question is whether companies will trust browser-based agents with enough permissions to save real time without creating new risk. Google’s answer this week was to make Chrome and Gemini look less like separate tools and more like one managed workplace surface. (cloud.google.com) (cloud.google.com)