EN ▾
SNAPE REP RT Front page →
ConfirmedAI safety05 Nov, 13:00 UTC

Anthropic leaves Claude Dispatch flaw open that lets a browser extension run code through Claude Desktop

Photo of a screen showing the Claude logo, used by The Register for its report on the Claude Dispatch flaw.
The Register

Forever Security's Gal Weizman told The Register the 'ThisBash' issue was reported in May; Anthropic marked it a duplicate, paid no bounty, issued no CVE and said it planned to retire Dispatch.

Read at The Register Share

Why it matters

Agent features turn ordinary browser trust into code execution: Weizman found Dispatch's interface reachable from a normal logged-in claude.ai session, so any extension able to control that page could drive Claude Desktop on the same machine or another signed into the account. The Register says Anthropic has known since May, closed the report as a duplicate and planned to retire Dispatch at the end of October, yet its support page still listed the feature on 30 October. Watch for a fix, a retirement date or a CVE.

Key facts

  • Flaw dubbed 'ThisBash', found by Gal Weizman of Forever Security
  • Dispatch reachable from an ordinary logged-in claude.ai session
  • Can reach another machine signed into the same Claude account
  • Reported in May; closed as a duplicate
  • No bug bounty and no CVE issued
  • Support page on 30 October still listed Dispatch in limited beta

Sources

More on AI safety
More from Snape ReportFront page →