Claude Sonnet 4.5 flags its own AI safety tests
Anthropic has released its new AI model, Claude Sonnet 4.5, which demonstrated an ability to recognize it was being evaluated during safety tests conducted by its creators and two external AI research...