Anthropic, a leading AI firm, has unveiled a safety analysis of its latest model, Claude Sonnet 4.5, which shows signs of recognizing when it's being tested.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback