Advancing AI in Government: A New Blueprint for Responsible Acquisition
In April 2025, the Office of Management and Budget (OMB) issued Memorandum M-25-22, authored by Director Russell T. Vought, to overhaul how federal agencies acquire artificial intelligence (AI) systems. This new directive, titled “Driving Efficient Acquisition of Artificial Intelligence in Government,” reflects the evolving policy landscape shaped by recent executive orders and legislative mandates, including the Advancing American AI Act and Executive Orders 13960 and 14179. It replaces prior guidance (M-24-18) and aims to ensure that the federal government embraces AI in a way that is safe, secure, innovative, and ultimately beneficial to the American public.
At the heart of the memorandum are three overarching principles: cultivating a competitive American AI marketplace, safeguarding taxpayer dollars through rigorous performance tracking and risk management, and fostering cross-functional collaboration throughout the AI acquisition lifecycle. Agencies are encouraged to avoid vendor lock-in by prioritizing data portability and interoperability, and to ensure that acquisition requirements support open and transparent AI solutions. The guidance recognizes that AI has the power to enhance government operations, but only if it is procured and managed responsibly.
The memorandum’s implications are far-reaching. For one, it applies to all federal agencies covered under 44 U.S.C. § 3502(1), except for elements of the Intelligence Community and national security systems. It mandates that future AI contracts—whether newly awarded or renewed—comply with its standards within 180 days of issuance. Agencies must also update internal acquisition policies within 270 days to reflect the new guidance and ensure adherence to existing rules like those in Memorandum M-25-21.
One of the most critical updates involves how AI-related intellectual property (IP) rights and government data usage are handled. Agencies must carefully delineate ownership rights, ensure government access to components needed to operate AI systems, and explicitly prohibit vendors from using agency data to train commercial AI models unless consent is given. Additionally, privacy protections must be embedded in every contract, with special attention paid to personally identifiable information.
The memo also spotlights performance-based acquisition techniques such as using Statements of Objectives, Quality Assurance Surveillance Plans, and tailored contract incentives to evaluate and improve vendor performance. It stresses the importance of transparency and requires that agencies provide vendors with detailed documentation needs, particularly when dealing with high-impact AI use cases. To support implementation, the General Services Administration (GSA) and OMB will publish procurement guides and create a secure repository for agencies to share tools, contract language, and best practices.
Ultimately, M-25-22 signals a pivotal shift in how the federal government approaches AI—balancing the promise of innovation with the imperative for security, ethics, and accountability. By laying a structured framework for acquisition, the memo not only promotes efficiency and fairness but also builds the public trust necessary for widespread adoption of AI in public services.
This blog post is a summary for informational purposes only. It is not guaranteed to be accurate, complete, or up to date, and does not constitute legal advice.
