February 14, 2024 — As noted in a recent letter to the National Institute of Standards and Technology (NIST) from the House Committee on Science, Space, and Technology, AI safety tools and methodologies and claims about AI technologies themselves are contested and lack scientific validity. Accordingly, NIST has the important task of building consensus across the AI ecosystem to construct empirically sound, rights-protecting AI safety methods that can be widely adopted.

In response to the request for information related to NIST’s assignments under sections 4.1, 4.5 and 11 of the Biden-Harris administration’s executive order concerning artificial intelligence (EO 14110), we argue that the process of achieving consensus to build AI safety and security practices must be participatory and include civil society and historically impacted communities. The process must also be critical and precise about the problems that methods and standards can address.