Claude Cowork Now Has Its Own Built-In Browser

Anthropic has introduced an integrated browser feature to Claude Cowork, enabling the AI assistant to access the web independently of the Claude in Chrome extension. The built-in browser automatically launches in a side panel when a prompt necessitates web-based information. This feature is enabled by default and currently accessible to Enterprise customers, with Pro, Max, and Team subscribers set to receive the update this week. Users have the option to revert to the Chrome extension via settings.

Anthropic emphasizes that the browser's scope is deliberately limited. It cannot interact with logged-in services like email or banking accounts without users inputting their credentials manually. The primary function of the in-app browser is to fetch public information, not to execute actions through a user's existing browser sessions.

This development is significant given that Chrome accounts for nearly 70% of global web traffic, as per Statcounter. Additionally, European efforts to diversify browser options under the Digital Markets Act do not apply to in-app browsers, as they are exempt from the requirement to display a choice screen during initial use.

Anthropic cautions that while the new browser incorporates security guardrails similar to those in the Claude extension, it does not eliminate the risk of prompt injection. This form of attack involves concealing malicious instructions within webpages, potentially manipulating AI agents into carrying out unintended actions. Just as OpenAI has integrated its standalone browser into a Work mode, Anthropic is increasingly embedding browser functionalities within AI assistants.

The trend suggests an evolving landscape where users might interact more frequently with embedded browsers rather than traditional standalone browsers.