OMB Releases a Pair of Memos Reflecting Trump AI Priorities

Upon taking office in January, the Trump Administration promptly fulfilled a campaign promise to repeal President Biden’s AI Executive Order (EO). Shortly afterwards, the Trump Administration released EO 14179, Removing Barriers to American Leadership in Artificial Intelligence . The EO directs the review and revision of the previous administration’s OMB memos on AI technologies. In response, OMB issued a pair of memorandums to agencies on April 3rd on the federal use and procurement of AI technologies.

Principal Deputy Director of the White House Office of Science and Technology Policy (OSTP), Lynne Parker, stated that the revised memos will streamline AI adoption and procurement by removing bureaucratic restrictions, enhancing efficiency and cost-effectiveness, and supporting a competitive American AI marketplace.

Details on each memo are below.

M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust

Description

This memo rescinds and replaces M-24-10 and charges agencies with removing barriers to federal AI innovation, pursuing efficiencies in AI adoption, and installing a governance and risk management framework to promote trustworthy federal use of AI.

Action dates (from date of memo release)

• After 60 days, requires agencies to retain or designate a Chief Artificial Intelligence Officer (CAIO)
• After 90 days, requires Chief Financial Officers (CFO) Act agencies to coordinate issues within an agency AI Governance Board
• After 90 days, requires OMB to convene a CAIO Council
• After 180 days, requires CFO Act agencies to release AI strategies focused on removing barriers to AI use and maturity
• After 180 days, requires agency compliance plan to the memorandum
• After 270 days, requires updated internal policies on IT infrastructure, data, cybersecurity and privacy
• After 270 days, requires agencies to develop Generative AI policies
• After 365 days, requires agencies to implement minimum risk management practices for high-impact uses of AI (excludes elements of the IC)
• After 365 days, requires agencies to publicly report determination and waivers for AI use cases after reporting directly to OMB on an annual basis (excludes elements of the IC)
• On an annual basis, requires agencies to publish and update AI use case inventories (excludes elements of IC and the DOD)

Thoughts for Contractors

• Governance Infrastructure Largely Remains the Same. Designation of agency CAIOs and governance boards, as well as a government-wide CAIO council are not new concepts. However, according to the memo’s fact sheet, CAIO roles are “redefined to serve as change agents and AI advocates, rather than overseeing layers of bureaucracy.”
• Efficiency in AI Adoption. The memo directs agencies maximize the value of existing investments to ensure speedy deployment as well as reuse resources for AI adoption, including agency data, models, code and performance assessments. Agencies are also instructed to release and maintain AI code as open-source software in a public repository.
• Agency Assessment of AI Maturity Goals. Agencies are directed to ensure adequate access to IT infrastructure and quality data, as well as access to software tools and open-source libraries to accelerate development, testing and maintenance of AI applications.
• Improved Performance Tracking. Agencies must optimize tracking and evaluation of acquired AI using methods such as documenting known capabilities, assessing for overfitting to known test data, and considering contractual terms that prioritize continuous improvement and performance monitoring.
• Risk Management of High Impact Systems. The definition of “high-impact” systems closely aligns to that of “safety-impacting” or “rights-impacting” systems referred to in the Biden Era. Nonetheless, the AI memo describes the risk management framework for such systems using language such as “minimum” practices. Additionally, the fact sheet states that accountability of such systems should mirror existing government IT processes vs. creating new approval layers.

M-25-22, Driving Efficient Acquisition of Artificial Intelligence in Government

Description

This memo rescinds and replaces M-24-18 and is founded on three key principles for AI procurement: ensuring government and public benefit, safeguarding taxpayer dollars, and promoting effective AI acquisition.

Action dates (from date of memo release)

• After 100 days, requires GSA to develop a public guide to assist agencies with AI procurement
• After 180 days, expects agency compliance with memorandum’s guidance in issued solicitations and any options to renew or extend a current contract
• After 200 days, requires agencies to standardize data ownership and IP rights processes in AI procurement
• After 200 days, requires GSA to develop a web-based repository of tools and resources to enable AI procurement

Thoughts for Contractors

• Pay Attention to the Memo’s Acquisition Lifecycle Guidance. The memo dedicates 4+ pages to AI in the acquisition lifecycle, from Identification of Requirements to Contract Closeout. Eye-catching elements include:
  • Prompting agencies to seek out innovative solutions from new companies that have not previously done business with the federal government.
  • Encouraging detailed demonstrations and tests of AI systems for market research.
  • Supporting performance-based acquisition techniques in requirements and contract terms.
  • Considering testing environments of proposed solutions on government systems during proposal evaluation.
  • Promoting the use of language to avoid vendor lock-in within solicitation, evaluation and contract closeout.
• Emphasis on American-Made AI. Both memos strongly encourage agencies to invest and maximize the use of American-made AI products and services.
• Emphasis on Avoiding Vendor Lock-in. On more than one occasion, the memo directs agency awareness to avoid cost dependencies on a single vendor by implementing clear and concise contractual requirements on the transfer of data, properly delineating data ownership and Intellectual Property (IP) rights for the government and contractor, and using product evaluations to identify the cost-effectiveness of a solution.

Concluding Thoughts

With the memos providing a solid glimpse into the AI framework of the Trump Administration, industry and academia await to see how requirements from both documents are implemented. Looking ahead, the White House also directed the development of an AI action plan within 180 days of EO 14179’s release, placing that deadline toward the end of July 2025. NSF subsequently issued a Request for Information (RFI) welcoming comments on the action plan with responses due last month.

AI enablement is also progressing under the current administration. Last week, Energy announced the identification of 16 federal sites for data center and AI infrastructure development alongside an RFI release that is soliciting comments on the “potential development approaches, technology solutions, operational models, and economic considerations associated with establishing AI infrastructure on DOE sites,” according to the RFI.

To follow the movements of the new administration, refer to GovWin’s First 100 Days Resource Center.