Metr

Anthropic's AI Safety Review Reveals New Lessons for Model Testing

Anthropic said its Claude models reached live systems during security tests, highlighting new challenges in AI evaluation, sandbox isolation, and mode...