FYI: Science Policy News
FYI
/
Article

Government-wide Policy on AI Issued by White House Budget Office

APR 01, 2024
The policy establishes many security and safety requirements for federal applications of AI but includes exemptions for basic and applied research.
Jacob Taylor headshot
Senior Editor for Science Policy, FYI American Institute of Physics
Vice President Kamala Harris attends a meeting with Artificial Intelligence (AI) CEO’s, Cabinet members and senior advisers, Thursday, May 4, 2023, in the Roosevelt Room of the White House. CEO’s in attendance are Sam Altman of OpenAI, Dario Amodei of Anthropic, Satya Nadella of Microsoft, and Sundar Pichai of Alphabet.

Vice President Kamala Harris, pictured speaking with AI company executives in May 2023, announced the White House’s policy on uses of AI across government in a speech on March 28, 2024.

(Lawrence Jackson / The White House)

Vice President Kamala Harris announced a major policy last week that will govern how federal agencies can use artificial intelligence, as directed by President Joe Biden’s November executive order on AI.

The policy, issued by the White House Office of Management and Budget, establishes many security and safety requirements for federal applications of AI but also includes various exceptions. Notably, the policy does not apply to AI used to conduct basic or applied research unless the purpose of that research is to develop AI applications for use by the agency.

The policy defines a set of “minimum practices” that agencies are required to follow when using or creating rights-affecting or safety-impacting AI – that is, AI designed to inform decisions that affect the rights or safety of individuals.

These practices include real-world testing, independent evaluation, public documentation, impact assessments, and other basic software risk-mitigation standards. Agencies are also required to “proactively” share AI-related code they develop, and the policy specifically recommends doing so via the National AI Research Resource.

Agencies have until December 1 to make any AI they are currently using compliant with the policy or halt their use. Agencies can request waivers to allow them to continue using non-compliant AI. Among the remaining provisions, agencies are required to annually update their inventories of AI use-cases and designate chief AI officers.

This news brief originally appeared in FYI’s newsletter for the week of April 1.

Related Topics
More from FYI
FYI
/
Article
Kevin Geiss will lead the arm of the Air Force Research Lab that focuses on fundamental research.
FYI
/
Article
An NSF-commissioned report argues for the U.S. to build a new observatory to keep up with the planned Einstein Telescope in Europe.
FYI
/
Article
Space, fusion energy, AI, quantum technology, and semiconductors were among the topics of discussion.
FYI
/
Article
The camera has a lens that is more than five feet across and will be installed at the Rubin Observatory in Chile.

Related Organizations