Anthropic has reduced false positives in its biology safety filters for Fable 5 by approximately 85 percent. Previously, nearly all biology queries were blocked and rerouted to the less capable Opus 5, drawing sharp criticism from scientists. Users can now handle far more biology tasks with Fable 5, such as interpreting lab results, understanding symptoms, and answering medical questions. Restrictions remain for dual-use topics like virology, toxicology, and molecular design, where Anthropic says Fable 5 could give bad actors capabilities unavailable elsewhere. The company is also building access programs so researchers can eventually use these restricted features.
The company argues biological risks may be harder to contain than cyberattacks because a released virus cannot be shut down and countermeasures take time. An analysis by U.S. intelligence agencies, a review of ChatGPT chats requesting bioweapon instructions, joint warnings from researchers, and a Stanford project that used AI to design fully synthetic viruses all suggest the threat is real and not just the usual Anthropic angst. This approach balances usability for legitimate science with caution on high-risk applications.
* False positive rate for general biology queries dropped significantly
* Dual-use topics remain restricted for public access
* Special access programs are being developed for qualified researchers
