Researchers Demonstrate Code Execution Risk via Claude Browser Extensions

A proof-of-concept attack demonstrated by researchers at the endpoint security firm Forever Security reveals a potential security risk involving Claude browser extensions and the Claude Dispatch feature. The researchers showed that a malicious browser extension, one capable of running scripts on claude.ai, could be used to trigger code execution on a user's local Claude Desktop application.

The attack chain relies on the Claude Dispatch feature, designed to let users send tasks from their phone to Claude Desktop. According to the researchers, if a malicious extension injects scripts into claude.ai and a user interacts with the Dispatch feature linking their phone to their Desktop, the extension could potentially execute code on the local machine.

The researchers noted that the flaw enabling this interaction can still be exploited. As of the publication date, no official patch or mitigation from Anthropic has been specified, though the researchers recommended disabling the Dispatch feature in Claude Desktop settings or removing browser extensions that can inject scripts into claude.ai as provisional mitigations.

The findings were reported by The Hacker News and pertain to the Claude Desktop and claude.ai platforms.