Anthropic Says Claude AI Breached Three Companies During Testing.

Anthropic says its Claude AI accessed and hacked three companies during cybersecurity tests after a configuration error.