A New AI Frontier: Claude’s Chrome Extension Opens the Door to New Dangers

On 27 August 2025, Anthropic released a Chrome-based extension for Claude Max users that reads active web pages and executes tasks based on user instructions, currently available only as a limited research preview. Once installed, the extension automatically activates the browser’s sidebar, letting users interact with Claude on the right while continuing to browse on the left.

Anthropic warns that using AI in the browser raises the risk of prompt injection attacks, where malicious content embedded in websites, emails, or documents could trick the AI into carrying out dangerous actions without the user’s knowledge. Such attacks could potentially lead to file deletion, data theft, or unauthorised financial transactions, according to Anthropic’s own red-team testing, which they say demonstrates these threats are feasible.

For security reasons, Claude is blocked from accessing banking and financial services, investment platforms, adult sites, and cryptocurrency exchanges, though Anthropic admits it cannot anticipate every domain and invites users to report unblocked sites.