Claude's collaboration tools allowed remote code execution
go.theregister.com
: Anthropic fixed the flaws - but the AI-enabled attack surfaces remain