Anthropic Updates Claude Fable5 Model, Slashing Biology False Positives by 85%

U.S. AI startup Anthropic announced yesterday an update to the biosafety safeguards of its Claude Fable5 model. The update refines the safety classifier’s mechanism, enabling it to more accurately distinguish between harmless queries and high-risk ones. According to official test data, the updated Fable5 shows an 85% reduction in performance degradation when handling biology-related questions. Anthropic stated that the safety classifier is designed to prevent users from exploiting the model for high-risk activities such as biological weapons research; therefore, Fable5 will continue to restrict requests related to specialized biological research and drug development.