Anthropic has updated Claude Fable 5’s biology safeguards to substantially reduce false positives, resulting in an approximately 85% decrease in biology-related fallbacks across its product surfaces. This change allows the model to assist with a wider range of everyday health and educational queries while still blocking dual-use professional research.
- The update rewrites the safety classifier’s constitution to better distinguish between benign uses and harmful dual-use content like virology or toxicology.
- Fallbacks are reduced by roughly 67% on Claude.ai, 55% on Cowork, 17% on Claude Code, and 7% on the Claude Platform.
- Healthcare professionals will receive more support on clinical tasks, though professional biology research and drug development remain blocked.
The company aims to balance the catastrophic risks of biological weapon development with the need to provide frontier AI capabilities to legitimate researchers through trusted access pathways.