Anthropic Claude in Chrome goes generally available with autonomous browser actions

On August 26, Anthropic announced that Claude in Chrome is generally available on all paid Claude plans. Users can now ask Claude to take actions in the browser, including reading and typing text, clicking links, navigating between pages, and filling out forms, using their existing logins.

A key upgrade is that Claude can now take actions autonomously in the browser, without requiring approval for each step. A safety classifier validates each action before it is performed to ensure it is safe and matches the user's original request. If the action does not match the request, it is blocked.

Anthropic also published evaluation results for its defenses against prompt injection attacks. In initial evaluations using the Cowork harness, no attack successfully bypassed the defenses of Claude Fable 5, Claude Opus 5, or Claude Sonnet 5. Claude is trained against a growing library of prompt injection attacks sourced from internal automated attackers, external red teams, and real-world monitoring. When a new attack succeeds against a current model, it is added to the library to inform future models and deployed safeguards.

Anthropic first published its prompt injection defenses for browser use in November 2025. Since then, the team improved how it trains both the model and its probes, and added an additional set of classifiers that make it possible for Claude to safely take more autonomous actions in Chrome. In Chrome, Claude will now automatically approve actions it determines to be safe, using the same mechanism as auto mode in Claude Code.

Claude in Chrome works across tabs and syncs with Claude's desktop, mobile, and web apps. This allows Claude to access internal dashboards, legacy systems, and vendor portals that lack native integrations.