On August 7, local time, artificial intelligence company Anthropic said it had updated Claude Fable 5's biosafety mechanism to significantly reduce erroneous interceptions by optimizing the classifier. In testing, this update reduced the number of biological-related “downgrade switches” across product platforms by about 85%, and users can use Fable 5 to handle more biology-related tasks. Anthropic said this adjustment will reduce erroneous interception in scenarios such as routine health consultation, interpretation of experimental results, and biological education, while providing more clinical task support for medical professionals. However, for specialized biological research and drug development requests with “dual-use” risks, such as virology, toxicology, and molecular design, Fable 5 will still switch to a less capable model to handle.

On August 7, local time, artificial intelligence company Anthropic said it had updated Claude Fable 5's biosafety mechanism to significantly reduce erroneous interceptions by optimizing the classifier. In testing, this update reduced the number of biological-related “downgrade switches” across product platforms by about 85%, and users can use Fable 5 to handle more biology-related tasks. Anthropic said this adjustment will reduce erroneous interception in scenarios such as routine health consultation, interpretation of experimental results, and biological education, while providing more clinical task support for medical professionals. However, for specialized biological research and drug development requests with “dual-use” risks, such as virology, toxicology, and molecular design, Fable 5 will still switch to a less capable model to handle.