Improving Fable 5's biology safeguards

Anthropic is updating Claude Fable 5's biology safeguards to cut down on false positives. In testing, this reduces biology-related fallbacks by about 85% across product surfaces, so users will see far fewer instances where the system switches to a less capable model after a biology query. This allows Fable 5 to assist with a wider range of biology tasks, including everyday health and educational questions like interpreting lab results, understanding symptoms, and learning biology, as well as giving healthcare professionals more support on clinical tasks.

However, Fable 5 still falls back to Opus 5 for dual-use requests in virology, toxicology, and molecular design, so it is not yet usable for professional biology research and drug development. Anthropic says it is committed to closing that gap through trusted access pathways, believing the greatest opportunity for AI is in biology and medicine.

The post explains why the safeguards exist: Fable 5 can now outperform experts on some highly complex biological tasks and offer operational support on others, which could help researchers develop treatments but also be misused by malicious actors, for example to develop biological weapons. Anthropic's capability assessments show Fable 5 could provide significant uplift to such actors—capabilities they could not find elsewhere.

Anthropic notes it is often hard to distinguish beneficial and harmful uses: researching a disease treatment may require producing dangerous compounds, as with live vaccines requiring scientists to grow the same pathogen, or the hypertension drug captopril, which was developed from toxic snake venom components that lower blood pressure. As frontier AI biology capabilities develop, the company says it needs to be cautious to ensure new risks don't materialize ahead of potential scientific benefits.