Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government
This is an executive order issued by President Donald J. Trump on December 3, 2020, directing federal agencies to promote the responsible use of artificial intelligence (AI) in government. It establishes a core set of nine AI Principles agencies must follow when designing, developing, acquiring, and using AI for non-defense, non-intelligence missions, with an emphasis on public trust, privacy, civil rights, and civil liberties. The order also creates a framework for implementing these Principles through guidance from the Office of Management and Budget (OMB), mandatory inventor ies of AI use cases, interagency coordination, and workforce development initiatives to build AI expertise across the government. Excluded from the order are AI used by the Department of Defense and the Intelligence Community, as well as AI embedded in common commercial products and certain AI research activities. The overall aim is to balance AI innovation with safeguards and accountability, ensuring that AI serves the public interest and remains consistent with laws and American values. Put simply, the order requires agencies to (1) follow the Principles for trustworthy AI, (2) follow a coordinated implementation path guided by OMB, (3) inventory and publicly disclose how AI is being used across agencies, and (4) invest in workforce programs to grow AI skills within the federal government, all while excluding defense/intelligence applications and certain other narrow exceptions.
Key Points
- 1Principles for Use of AI in Government: Agencies must design, develop, acquire, and use AI in a way that is lawful, purpose-driven, accurate, safe, understandable, responsible and traceable, regularly monitored, transparent, and accountable, with appropriate training for personnel.
- 2Policy Guidance and Standards: OMB is charged with issuing common policy guidance to implement the Principles, including leveraging voluntary consensus standards where feasible and updating guidance as AI use evolves.
- 3AI Use-Case Inventories and Public Disclosure: Agencies must (a) create criteria and guidance for reporting non-classified, non-sensitive AI use cases, (b) regularly inventory these use cases, assess for consistency with the order, (c) develop plans to align or retire misaligned AI, and (d) share inventories with other agencies and, within 120 days, make them publicly accessible where practicable.
- 4Interagency Coordination and Workforce Development: The order directs interagency collaboration, including publishing a list of coordination bodies; it also creates programs to build AI expertise (Presidential Innovation Fellows track and government rotational programs) and requires OPM to report on how rotations can expand AI talent.
- 5Scope and Exclusions: The order applies to non-defense, non-intelligence agency uses of AI that support missions, decision making, or public benefits, and allows exemptions for AI used in defense/national security, AI embedded in common commercial products, and AI R&D activities. It defines the scope with reference to NDAA 2019 for AI and clarifies that existing laws and budgets govern implementation.