On August 7, 2026, Anthropic announced an update to the safety mechanisms on its top-tier AI model, "Claude Fable 5," in the field of biology, reducing excessive restrictions (fallbacks) on routine health and education-related queries by approximately 85%. This addresses an issue where safety classifiers mistakenly blocked benign questions, significantly improving usability for healthcare professionals and general users. However, specialized dual-use research applications, such as virology and molecular design, remain restricted, meaning the model cannot be used for full-scale drug discovery. The company is developing a "trusted access pathway" for researchers, but a specific launch timeline has not been set. According to the company's verification, the incidence of "fallbacks"—where the model automatically switches to the less capable "Claude Opus 5" for biology-related queries—has decreased by approximately 85% across the product. With this update, the advanced reasoning capabilities of Claude Fable 5 can now be utilized for a wider range of biology-related tasks, including everyday health consultations and educational purposes. In an official blog post, Anthropic expressed the view that "the greatest opportunity for AI to have a positive impact on the world lies in biology and medicine," while explaining that it is investing heavily in building "responsible methods" to provide researchers with access to cutting-edge biological capabilities. Due to its extremely high performance, Claude Fable 5 had been flagged for the risk of misuse by malicious actors for purposes such as