GSA's Proposed AI Procurement Rules: What Federal Contractors and Agencies Need to Know

On March 6, 2026, GSA released a proposed contract clause — GSAR 552.239-7001 — that would impose sweeping new requirements on federal contractors using AI in contract performance. The proposal was deferred from MAS Refresh 31 to Refresh 32 after significant industry pushback, with the comment period extended to April 3, 2026. The rule is not yet final, but its direction is clear enough that both agencies and vendors on the MAS schedule should be preparing now.

This is not a routine regulatory update. The proposed clause would change fundamental aspects of how AI-enabled software is procured, used, and owned under federal contracts — including provisions that go significantly further than most vendors have anticipated.

The Three Core Requirements

Why the Industry Pushed Back

The comment period extension and deferral from MAS Refresh 31 reflect substantial industry concern about all three provisions — but particularly the third.

The "Custom Developments" ownership provision has the widest potential scope. If a contractor uses a foundation model and fine-tunes it on agency-specific data to improve performance for that contract, the resulting fine-tuned model would be government property under the proposed rule. For AI vendors whose business model involves continuously improving their models from deployment experience, this creates a fundamental tension: the improvements that make the product better for one agency become government property, unavailable for use with other customers.

The disclosure requirement also drew criticism for its breadth. Requiring disclosure of all AI tools used in contract performance — including commercial productivity tools like AI-assisted writing tools that employees use incidentally — creates a compliance burden that scales poorly for large organizations with thousands of contracts.

What This Means for Federal Agencies

For agencies buying AI-enabled software, the proposed rule has different implications than for vendors. If finalized, it creates several practical considerations:

• Pre-award disclosure requirements — agencies will need processes to collect and review AI disclosure information within the 30-day window, which requires administrative capacity that most don't currently have built into their onboarding processes
• American AI verification — contracting officers will need to verify that AI tools used in performance meet the "American AI Systems" standard, which requires understanding of the underlying technology that most acquisition teams currently lack
• Custom Development ownership tracking — if the government owns fine-tuned models, agencies need processes to identify, document, and potentially leverage those assets — which adds complexity to contract administration
• Solicitation updates — even though the rule isn't final, agencies with upcoming AI-related acquisitions should consider whether to incorporate these requirements proactively, both for compliance preparation and to signal to vendors that this is the direction of travel

How to Prepare Before Finalization

The rule is proposed, not final. It may change significantly before finalization, or may not be finalized at all. But the underlying policy direction — greater transparency around AI use, preference for domestic AI systems, government rights in AI developed with government data — is clearly established by the existing OMB framework and unlikely to reverse.

For agencies, the practical steps are:

• Review current MAS contracts for AI-enabled services and identify which would be affected by the disclosure and ownership provisions
• Build AI disclosure collection into new contract onboarding processes now, before the rule requires it
• Brief your legal counsel and contracting officers on the "Custom Developments" definition, since contract-by-contract interpretation will be required when the rule finalizes
• For new AI acquisitions, include provisions in solicitations that address AI transparency and domestic-origin requirements, even without a final rule

The GSA AI procurement rule represents the federal government's first serious attempt to regulate how AI is used in contract performance — not just how it's acquired. Whether GSAR 552.239-7001 is finalized as proposed, modified substantially, or withdrawn, the questions it raises about AI transparency, domestic sourcing, and government rights in AI models are not going away. Agencies and vendors that build answers to those questions into their processes now will be better positioned than those that wait for a final rule.