Anthropic Brings Claude to Chrome in Early Preview

Anthropic today released a Claude extension for Google Chrome in an early pilot program limited to select Max subscribers.

“We’ve spent recent months connecting Claude to your calendar, documents, and many other pieces of software,” the Anthropic announcement explains. “The next logical step is letting Claude work directly in your browser.”

While Anthropic feels that AI in the web browser is inevitable, it’s also taking note of recent reports about security vulnerabilities associated with browser-based AI agents. And so this initial test of its first Chrome browser extension is about getting real-world feedback from trusted partners. And then Anthropic will expand access as it develops safety measures.

“Browser-using AIs face prompt injection attacks where malicious actors hide instructions in websites, emails, or documents to trick AIs into harmful actions without users’ knowledge,” Anthropic says. “Prompt injection attacks can cause AIs to delete files, steal data, or make financial transactions. This isn’t speculation: we’ve run ‘red-teaming’ experiments to test Claude for Chrome and, without mitigations, we’ve found some concerning results.”

This seems like a curious way to advertise a new offering, but Anthropic says it has already implemented several defenses that reduce the attack surface. And it’s building additional safeguards into the extension based on its trustworthy agent principles. Among these safeguards, the extension isn’t available when browsing financial services sites or those with adult or pirated content.

If you’d like to take part in this early testing—and with a description like that above, who wouldn’t?—you can join the Claude for Chrome research preview waitlist.