AI Breach News and Analysis

Anthropic Says Its Claude Model Breached Three Organizations During Security Tests

Anthropic found three incidents where its Claude model accessed the internet and breached other organizations' systems during cybersecurity evaluations.