The federal government’s new AI playbook asks agencies to move faster while preserving the ability to stop systems that fail. That balance is the central story in OMB Memorandum M-25-21: adoption is encouraged, but consequential uses must carry visible owners, documented safeguards, continuing review, and a path to discontinuation.
Background
On April 3, 2025, OMB Director Russell T. Vought issued M-25-21, “Accelerating Federal Use of AI through Innovation, Governance, and Public Trust.” The memorandum rescinds and replaces M-24-10 and implements requirements associated with Executive Order 14179, signed January 23, 2025. It applies across Executive Branch agencies, including independent regulatory agencies, with specified exceptions for national security systems and certain research, regulatory, and enforcement-related activities. The OMB memorandum organizes the policy around three priorities: innovation, governance, and public trust.
The innovation agenda is direct. Agencies are told to remove unnecessary bureaucratic barriers, improve public services, increase efficiency, reuse existing investments, and share data, code, models, and assessments where practicable and legally permissible. When acquiring AI, they should maximize the use of products and services developed and produced in the United States. They are also encouraged to protect government data rights, support competition, and avoid vendor lock-in. These procurement provisions are recommendations in the memorandum, distinct from its mandatory requirements.
The memorandum’s safeguards turn on consequence rather than technical complexity. High-impact AI is AI whose output serves as a principal basis for decisions or actions with a legal, material, binding, or significant effect on civil rights, civil liberties, privacy, access to programs or services, health and safety, critical infrastructure, public safety, or strategic assets. Human oversight does not by itself remove a use case from that category.