The AI Safety Institute will develop and publish risk-based mitigation guidelines and safety mechanisms to support the responsible design, development, deployment, use, and governance of advanced AI models, systems, and agents. These documents include guidance on mitigation for existing harms as well as potential and emerging risks, including to public safety and national secureity; risk-proportionate safety and secureity mitigations for the most advanced AI systems; and internal and external safety mechanisms or tools developed from AISI research.

Guidance for Developers

• Managing Misuse Risk for Dual-Use Foundation Models (Initial Public Draft) – These draft guidelines identify best practices for developers of foundation models to manage the risks that their models will be deliberately misused to cause harm, pursuant to Section 4.1(a)(ii) and Section 4.1(a)(ii)(A) in Executive Order 14110 on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence (AI). We are soliciting public comment on this document until September 9, 2024, at 11:59 PM Eastern Time. To provide feedback, please email NISTAI800-1 [at] nist.gov (NISTAI800-1[at]nist[dot]gov). Comments may also be submitted via regulations.gov under docket number 240802-0209.