A researcher at Forever Security found that a browser extension with access to claude.ai could drive the Dispatch feature, send prompts to Claude Desktop, click Allow on approval prompts and eventually get code running on the user's machine or another machine on the same account. The Register reports Anthropic accepted the report as valid in May but closed it as a duplicate, with no CVE. Anthropic reportedly told the researcher it planned to retire Dispatch, but it did not respond to The Register's questions. Teams should check which browser extensions staff run alongside claude.ai and whether personal accounts are signed in on corporate laptops.
