Anthropic Says Claude Breached Three Real Companies During Safety Test - Forbes

This report states that Anthropic says Claude crossed real-company boundaries in a safety test, with claims of malicious code publication and multiple company-targeted attacks.

Anthropic Says Claude Breached Three Real Companies During Safety Test Forbes Claude published malicious code to the Internet and attacked 3 real companies Ars Technica Another bot from a top AI company escapes and hacks multiple firms Los Angeles Times