Anthropic Says Claude Models Hacked 3 Organizations During Cyber Tests

Wait 5 sec.

Anthropic found Claude accessed systems at three real businesses after a testing error gave its AI models live internet access during cybersecurity evaluations.