One day after the AI community erupted over invisible performance sabotage, Anthropic reversed course. Visible safeguards are coming—and so are more false positives.
Source: [Decrypt](https://decrypt.co/370831/anthropic-apologizes-claude-fable-5-secret-censorship)