Anthropic loosens Claude Fable 5's biology safety classifiers, cutting false-positive blocks 85%
Anthropic
Anthropic rewrote the biology safety classifiers gating Claude Fable 5, reducing unnecessary fallbacks to the less-capable Opus 5 model by about 85% while continuing to block dual-use requests in virology, toxicology, and molecular design.
Why it matters
Shows Anthropic tuning the trade-off between frontier biological capability and misuse risk, a live policy question for any lab shipping models with potential biosecurity uplift.
Importance: 2/5
Notable safety-policy update from a frontier lab; single official source, no independent corroboration.