Anthropic announced a major update to its Claude Fable model, reducing unnecessary biology-related AI fallbacks by around 85 per cent while continuing to block requests linked to dangerous biological research. According to the Dario Amodei-led firm, these changes will reduce the number of times users are automatically switched to a less capable AI model when asking biology-related questions. Claude Fable 5 will now answer a wider range of health, education and biology queries without triggering safety filters, with users experiencing fewer “fallbacks” where the model hands a request over to Opus 5, a less capable model, after detecting biology-related content. Internal testing showed the update reduced these biology-related fallbacks by about 85 per cent, allowing direct responses for topics such as interpreting laboratory reports, understanding symptoms, learning biology concepts and other educational questions. Healthcare professionals may also receive improved assistance for certain clinical tasks. However, Claude Fable 5 will continue to redirect queries involving areas it considers "dual-use," including virology, toxicology and molecular design, because these topics could be used for both beneficial research and harmful purposes. Anthropic stated it introduced strict biology protections because the latest version of Claude Fable 5 has become more capable in handling complex biological tasks, and the same AI abilities that could support medical research could also be misused by malicious actors attempting to develop biological weapons. Internal assessments indicate the model could provide knowledge not easily available elsewhere, increasing potential risk.