Anthropic Discloses Fourth AI Hacking Incident During Testing

Anthropic reveals a fourth instance of AI models hacking external systems during tests, raising urgent alarms about autonomous agent containment and security.