GSA’s New Proposed “American AI” Clause for Schedule Contracts: What Contractors Need to Know

On March 6, 2026, GSA dropped a surprising new proposed GSAR clause it plans to roll into GSA Schedule contracts in the next MAS refresh (Refresh 31, expected late March or April). The clause—GSAR 552.239-7001, “Basic Safeguarding of Artificial Intelligence Systems”—is a big deal for any contractor selling through the Schedules program, and potentially for their subcontractors and “service providers,” too. 

GSA released the proposed clause via the Schedule refresh comment process—not through traditional notice-and-comment rulemaking. Thus, there is a very short window to submit feedback—GSA is accepting comments on the proposed clause only through March 20, 2026 (either in the “comments” section of the webpage or to the MAS Program Management Office email (maspmo@gsa.gov)).

This proposed clause builds on a series of laws, Executive Orders, internal policies, and agency guidance, including OMB Memo M-25-22 (Efficient Acquisition of AI in Government; we wrote about it here). It is the latest step in a multi-year federal push to regulate AI procurement and steer agencies toward “American AI Systems”—generally, AI developed and produced in the United States. The proposed GSAR clause would do precisely that—ensure that contractors only use American AI. But, of course, there’s more—the proposed clause also includes multiple compliance obligations for Schedule contractors.

The clause includes the following key compliance requirements for Schedule contractors:

1. Use only “American AI Systems” in the performance of any GSA Schedule orders
2. Disclose all AI systems used in performance (whether American or foreign)
3. Enable human oversight by government officials
4. Report any incident (as defined by FISMA) within 72 hours and provide daily status updates
5. Provide a mechanism for government feedback
6. Provide (upon government request) documentation demonstrating compliance with the clause, including system documentation consistent with the NIST AI Risk Management Framework

Significantly, the clause explicitly states the contractor is responsible for its AI “Service Provider(s),” with “Service Provider” defined as “an entity that directly or indirectly provides, operates, or licenses an AI system but is not a party to the contract. Service Providers may or may not be subcontractors.” (Emphasis added.) Therefore, it appears the substantive compliance obligations extend to indirect use of AI in developing a product or service used in performance of the contract.

Even when using American AI, the proposed GSAR clause includes material restrictions in how a contractor (and/or its Service Provider(s)) uses that AI. There’s a lot to unpack here, but below are some highlights:

• The Government owns all Custom Development of the AI System (defined to include modifications, customizations, configurations, or enhancements both to the AI System and any associated implementations or workflows). Practically, this means that the Federal Government is claiming ownership of both the contractor’s code written to implement the use of American AI, as well as any underlying modifications to the American AI Service Provider’s code, if done specifically for the Government under the GSA Schedule order.
• The contractor and Service Provider are prohibited from using Government Data (another defined term) to train or improve the AI/ML model. 
• The contractor and Service Provider are required to segregate Government Data from commercial customers, delete all Government Data upon the conclusion of the Federal contract, and provide a written certification of deletion.
• The contractor must make “commercial efforts” to ensure the Service Provider’s AI System follows “Unbiased AI Principles.” This provision may require contractors to renegotiate their commercial agreements with American AI Service Providers.
• The proposed clause requires that the license provided by the contractor to the Government permit the Government to use the AI System “for any lawful Government purpose.” 
• The contractor must notify the Government of certain changes to AI System(s) and/or Service Providers used in performance of the contract.
  The Government may suspend use of an AI System for “performance issues.”
• The contractor is liable for reasonable decommissioning costs if the Government elects to termination the contract for a failure to comply with the “Unbiased AI Principles.”

The proposed clause states it takes precedence over any other conflicting provision, including the contractor’s commercial terms (such as those in its Terms of Service or End User License Agreement). This is a departure from GSA’s approach over the past decade or so, where GSA has given contractors significant flexibility and made a concerted effort to match commercial software terms, where not prohibited by statute. However, note the proposed clause allows for the parties to “bilaterally supplement or revise” certain sections of the clause (including the IP restrictions)—contractors should be mindful of this and seek to obtain important revisions where possible. Unfortunately, that negotiation flexibility does not extend to the six compliance obligations highlighted above.

GSA is clearly signaling where it wants the Schedules program to go on AI. The question for contractors is whether they can operationalize these new requirements and prohibitions quickly—and whether the final clause will acknowledge how AI is actually built, licensed, and managed in modern software environments. To prepare, contractors should begin inventorying their AI use (both direct and indirect); confirm whether the AI tools they use may qualify as “American AI;” review and prepare for needed updates to subcontract/Service Provider terms; and build incident reporting, segregation/deletion, and documentation processes. And certainly don’t miss the opportunity to provide substantive comments prior to March 20.