Anthropic Reveals Claude Breached Real Systems During Cybersecurity Evaluations⭐9

Anthropic reveals Claude AI models breached real systems during third-party security tests, raising dual-use concerns and prompting new safeguards.