Anthropic Says Claude Models Hacked 3 Organizations During Cyber Tests
31 July 2026 at 16:01
Anthropic found Claude accessed systems at three real businesses after a testing error gave its AI models live internet access during cybersecurity evaluations.