Anthropic is currently working on a test to make Claude work directly in the Chrome browser.
AI company Anthropic has launched a test program for a Claude extension in Google Chrome. The AI agent will then perform actions directly in the browser. This new functionality will initially be available to 1,000 users of the Max subscription, which costs around $100 to $200 per month. Interested users can already join the waitlist. A potential wider rollout will only follow after the testing phase.
Claude in Chrome
The new Chrome extension allows Claude to perform tasks such as clicking, filling out forms, or navigating between pages. Users can then chat with Claude in the browser via a side panel that appears on the right.
The functionality will initially only be available to 1,000 users of the Max subscription. They can sign up via a waitlist. Through this testing phase, Anthropic aims to gain insights into practical applications, weaknesses, and security risks. A wider rollout will only follow after that.
AI is becoming increasingly integrated into browsers. OpenAI is reportedly working on its own AI web browser and Microsoft recently launched a preview of Edge Copilot Mode. The rise of AI-powered browsers is being further accelerated by Google’s ongoing antitrust case. The judge may decide that Google must sell its Chrome browser, which AI companies like Perplexity, which has already made a bid, would be very eager to see happen.
New Risks
At the same time, Anthropic emphasizes in its blog post that browser-based agents also bring new risks, such as prompt injection attacks. In these attacks, malicious actors try to use hidden instructions to induce AI systems to engage in unwanted behavior.
read also
Claude Gets Ability to Terminate Conversations in Case of Misuse
Anthropic emphasizes that the testing period should help discover unknown attack vectors in practice. Feedback from this pilot will be used to further refine the security systems, permission settings, and underlying AI models.