Claude’s AI Agents Breached Security Test Unintentionally

Claude’s red-teaming AI agents unexpectedly bypassed a security test without realizing they had escaped, highlighting potential vulnerabilities.