Ask Daiv
All storiesAI BRIEF

Anthropic reduces false positives in Claude Fable 5’s biology safeguards

Anthropic says an update to Claude Fable 5’s biology safety classifier reduced biology-related fallbacks by about 85% in its testing. The change should allow more everyday health, education and clinical questions to stay with the model instead of switching to Opus 5. Requests involving dual-use professional biology, including virology, toxicology and molecular design, will still fall back to the less capable model. These performance and safety claims come from Anthropic.

Anthropic says it has refined the biology safety classifier used with Claude Fable 5, reducing biology-related fallbacks by about 85% across its product surfaces in internal testing. A fallback occurs when a query is routed to Opus 5, which Anthropic describes as less capable in biology. The company says users should see fewer interruptions for everyday health questions, educational biology and some clinical tasks.

The change does not remove the higher-risk controls. Anthropic says requests involving dual-use professional biology—such as virology, toxicology and molecular design—will continue to fall back because Fable 5 may provide capabilities that could be misused. The classifier is intended to distinguish benign work from potentially harmful or ambiguous research, while remaining resistant to attempts to bypass it.

For Australian organisations, the update may make permitted biology-related workflows more usable, but it does not turn Claude into a clinical authority or confirm suitability for regulated work. Healthcare and life-science teams should define approved uses, retain qualified human review, protect sensitive data and test the system against their own safety and compliance requirements. The reported 85% reduction and safety assessment are Anthropic’s own findings.

Read the original article at Anthropic News
Anthropic reduces false positives in Claude Fable 5’s biology safeguards · Ask Daiv