Fable won’t answer basic biology questions
Anthropic's new Claude Fable 5 AI model refuses to answer simple biology questions due to overly conservative safety guardrails designed to prevent bioweapon development.Summary
Anthropic has released Claude Fable 5, a powerful 'Mythos-class' AI model, but it is currently restricted from answering basic biology questions that are well within its capabilities. Instead of providing answers, the model frequently deflects such inquiries to the older Claude Opus 4.8.
Anthropic spokesperson Paruul Maheshwary explained that these broad filters are intentional and deliberately conservative. The company aims to prevent malicious actors from using the model for dangerous biological research, such as creating bioweapons. While the company acknowledges that these safeguards result in 'false positives'—blocking harmless questions about cell biology or common medical conditions—they consider it a necessary tradeoff to ensure safe deployment.
While Fable shows more flexibility regarding chemistry and cybersecurity queries, its biology limitations remain stringent. Anthropic intends to refine these detection systems over time and eventually provide unrestricted access to the scientific community to foster legitimate biomedical research and drug discovery.
(Source：The Verge)