NewsBite analysis · AI, BioTech
The Decoder
Anthropic loosens Fable 5's biology restrictions but keeps the guardrails on for virology and toxicology
Anthropic has made significant adjustments to the biology safety filters of its AI model Fable 5, reducing false positives by approximately 85%. This change addresses criticism from scientists regarding the previous blocking of nearly all biology-related queries, which forced users to revert to a less capable model, Opus 5. Users can now tackle a wider range of biological inquiries, including interpreting lab results and addressing medical questions. However, the company has retained strict restrictions on topics deemed 'dual-use', such as virology and toxicology, to prevent misuse by potentially malicious actors.
MonitorConfidence: MediumShort term horizon
Published 7 Aug 2026, 18:10 UTC1 min read
Visit original article ↗Lead image — The Decoder
Summary
TL;DR
- False positives in Fable 5's biology filters reduced by 85%
- Enhanced user capabilities for biological queries
- Remaining restrictions on virology and toxicology topics
- Concerns over biological risks versus cyberattack risks
- Plans for access programs for restricted functionalities
Interesting facts
- Biological risks are considered more challenging to contain than cyberattacks.
- Previous AI models were criticized for overly restrictive biology queries.
- US intelligence agencies have raised alarms about AI's potential misuse in bioweapons.