AI-Bio Workstream

The Frontier Model Forum’s AI-Bio workstream aims to advance the biosafety of leading AI models and systems. Although frontier AI holds significant promise for medicine and the life sciences, it also risks amplifying existing biological threats and introducing novel ones. Understanding how to safely manage those risks is an urgent challenge.

The AI-Bio workstream was established to address that challenge. The workstream aims to develop shared understandings of AI-Bio threat models, safety evaluations, and mitigation measures, as well as common approaches to capability and risk thresholds in the biological domain. Effectively managing biosafety risks will require greater coordination around current scientifically-grounded best practices, as well as further research into new safety methods and approaches.  

The AI-Bio workstream draws heavily on the biosafety and biosecurity expertise of the Forum’s member firms, and collaboration with external scientific experts is also central to its efforts. From virology and microbiology to bioengineering and pandemic preparedness, the Forum actively seeks to engage with leading researchers and practitioners across the biological and life sciences, in addition to experts in AI and global security. By leveraging collective knowledge from academia, industry, and government, the AI-Bio workstream remains grounded in the latest scientific understanding across a diversity of sectors and disciplines.

While the Forum is committed to information-sharing and transparency, publications related to AI and biological threat creation can introduce significant information and attention hazards. The FMF will only publish materials here after careful review and deliberation by domain experts.

Recent Publications

Preliminary Taxonomy of AI-Bio Safety Evaluations. December 20, 2024.