The General Services Administration released a draft contract clause on March 6, 2026, that would give the U.S. government an irrevocable, royalty-free license to use any AI system developed under a federal contract. The proposal, GSAR 552.239-7001, also bans the use of government data to train AI models and requires contractors to source only AI systems developed entirely within the United States. If finalized, it would override many existing commercial licensing terms and fundamentally change how agencies acquire artificial intelligence services.
What the clause demands
The regulation targets AI systems procured through GSA Multiple Award Schedule contracts. Contractors must ensure that all AI components-from training data to deployment infrastructure-originate from the United States, with no foreign elements permitted anywhere in the supply chain. Subcontractors and downstream service providers fall under the same restriction. Companies are also expressly prohibited from using government data to train, retrain, or improve their models.
On the disclosure side, contractors must reveal all AI tools in use within 30 days of contract award. Incident reporting has a 72-hour deadline. The licensing provision gives the government broad rights to use the AI system for any lawful purpose, functionally making it government property without additional payment or permission.
Industry pushback and the pause button
Federal contractors and major AI vendors, including OpenAI and Anthropic, along with the Business Software Alliance, objected to the clause's scope. The concerns centered on the blanket prohibition on foreign sourcing, the override of commercial license terms, and the lack of exceptions for off-the-shelf products that include non-U.S. components. In response, the GSA extended the public comment period to April 3, 2026.
More significantly, the agency suspended the clause's inclusion in the upcoming MAS Refresh 31, a periodic update to Multiple Award Schedule terms. As of mid-June 2026, the proposal remains in draft, with possible consideration for Refresh 32.
The broader policy framework
The proposal fits within a broader administration strategy to standardize AI for Government procurement, as outlined in Office of Management and Budget memoranda M-25-21 and M-25-22. These directives emphasize using American-developed AI technologies that produce unbiased outputs and maintain strict supply-chain controls across federal agencies.
What this means for contracting officers
If GSAR 552.239-7001 is finalized, federal acquisition staff will need to verify that every AI system and its supply chain is American-made-a level of scrutiny far beyond typical IT security reviews. Contracting officers would also have to confirm that no government data feeds model training and that all AI tools are disclosed on time. The incident reporting deadline adds operational pressure for both agencies and vendors.
The licensing clause may cause some vendors to bypass the federal market entirely. Agencies should expect a smaller pool of qualified bidders and potentially higher prices for contracts involving AI development. Understanding these trade-offs will help acquisition teams plan ahead and communicate realistic expectations to industry partners.
Why this matters for government professionals
The clause's direction is clear even though it remains in draft. Government staff involved in AI acquisition should review the proposed language, submit comments, and begin assessing current contracts for exposure to these requirements. The shift toward American-built AI with strict data controls will demand new expertise in supply chain verification and intellectual property negotiation. Training resources like the AI Learning Path for Policy Makers can help policy and acquisition teams build the knowledge to navigate these emerging procurement rules.
Your membership also unlocks:
