On the Responsible Development and Use of Chem-Bio AI Models

This paper provides public comment in response to the Request for Information (RFI) by the AI Safety Institute (AISI), housed within the National Institute of Standards and Technology at the U.S. Department of Commerce, concerning the responsible development and use of chemical and/or biological (chem-bio) artificial intelligence (AI) models. The paper's purpose is to answer AISI's questions and share relevant expert knowledge on the subject areas prioritized in the RFI. Where possible, the authors have provided additional definitions and clarifications to help readers without a deep background in AI understand the key concepts, methodologies, and recommendations. 

This paper focuses on the responsible development and use of chem-bio AI models. Chem-bio AI models can help in analyzing, predicting, or generating novel chem-bio sequences, structures, or functions and are becoming increasingly capable and accessible for a wide range of dual-use applications. For these reasons, it is important to proactively address security and safety considerations, as the AISI RFI calls for. 

This paper aims to provide a detailed overview of the first four of five topics identified in the RFI. They are (1) dual-use risk assessment for chem-bio AI models, (2) frameworks for identifying, prioritizing, and implementing risk mitigations, (3) safety considerations when chem-bio AI models interact with other models and one another, and (4) impacts of chem-bio AI models on existing biodefense and biosecurity measures.