We believe the General Services Administration’s proposed AI procurement clause (GSAR 552.239-7001) risks pushing leading AI providers out of the federal market, ultimately leaving taxpayers with lower-quality services at higher costs. While we support the government’s goal of deploying AI responsibly, this current proposal threatens core intellectual property by forcing the exposure of trade secrets and authorizing overbroad government licenses that amount to ownership transfers. Furthermore, the clause actively undermines the Multiple Award Schedule framework by discarding established commercial norms and imposing unworkable flow-down requirements on prime contractors. Coupled with vague definitions and open-ended future compliance obligations, the proposal creates an unpredictable and legally risky environment for vendors. To ensure the government can continue to access world-class AI, we strongly urge the GSA to narrow its license grants, protect developers’ proprietary innovations and establish realistic transparency standards that do not dismantle the commercial framework.
NetChoice Comments for the Record RE: Proposed Government AI System Terms and Conditions (GSAR 552.239-7001)
April 3, 2026
NetChoice is a trade association committed to preserving free enterprise and free expression online. We engage on technology policy at the state, federal and international levels, and we welcome this opportunity to comment on the General Services Administration’s proposed AI procurement clause.
We recognize the government’s legitimate interest in deploying AI responsibly and applaud the effort to advance America’s AI industry and government efficiency through procurement. However, as written, GSAR 552.239-7001 would undermine these goals. The proposal threatens intellectual property rights, dismantles established contracting norms and introduces compliance requirements so vague they border on unenforceable. Without substantial revision, this clause risks pushing leading AI providers out of the federal market entirely, leaving taxpayers on the hook for lower-quality services at higher costs.
I. The Clause Threatens Core Intellectual Property
AI companies are built on proprietary models, training methodologies and system architectures. The proposed clause puts all of that at risk.
Forced Exposure of Trade Secrets. The clause mandates audit trails revealing intermediate processing steps, model routing logic and data retrieval methods. Developers guard such information as core trade secrets. Compelled disclosure would cause irreversible competitive harm. For prime contractors relying on third-party AI services, the problem is compounded: they simply cannot produce proprietary information they do not have access to.
Overbroad License Terms. Authorizing the government to use a proprietary AI system for “any lawful Government purpose” far exceeds what is necessary for contract performance. This language could allow federal agencies to deploy a contractor’s AI in unrelated programs, share access across agencies or effectively build competing capabilities without additional compensation or consent. The requirement to allow the government to “copy” and “store” AI systems compounds this concern, amounting to a transfer of ownership rather than a license.
Government Seizure of Custom Developments. The clause grants the government full ownership of any modifications, configurations or enhancements made during contract performance. This means a vendor could invest significant resources fine-tuning a model for a specific agency need, only to lose all rights to that work. Prohibiting contractors from applying their own expertise and innovations across other engagements would chill investment and deter the best AI developers from pursuing government work.
II. The Clause Undermines the Multiple Award Schedule Framework
The MAS program exists precisely to give agencies access to commercial products at commercial terms. This clause actively erodes that premise.
Discarding Commercial Norms. The draft clause explicitly overrides a contractor’s existing AI terms and commercial license agreements. It imposes national origin certification requirements, mandates data localization and requires deletion of data across all systems and backups. These obligations directly conflict with how commercial vendors manage data at scale. These aren’t minor adjustments; they require vendors to operate under an entirely separate set of rules for government customers.
Untenable Flow-Down Requirements. Prime contractors are held responsible for ensuring that third-party AI providers comply with the clause’s full requirements. This is unworkable. Prime contractors do not control, and often cannot access, the internal processes, architectures or data practices of their upstream AI vendors. Holding them legally responsible for compliance with obligations they cannot fulfill or verify creates significant legal exposure and will push many contractors to avoid AI-enabled offerings altogether.
III. Vague Definitions Create Unpredictable Obligations
Compliance requires clarity. This clause provides little of it.
An Undefined Scope. The clause’s definition of “AI System” is expansive enough to capture nearly any software with a machine learning component. While commercial products with “embedded AI” are nominally excluded, there is no standard for what qualifies as embedded. This ambiguity leaves contractors unable to determine whether their offerings fall within scope. The undefined term “AI components” creates similar uncertainty.
Unlimited Future Obligations. The clause requires contractors to implement future OMB guidance and directives as they are issued. This creates an open-ended compliance obligation tied to nonbinding agency documents that may not even exist at the time of contract award. Vendors cannot price, scope or manage risk around requirements that haven’t been written yet.
Opaque Assessment Standards. The government reserves the right to conduct automated assessments of contractor AI systems using undisclosed benchmarks. Contractors found non-compliant would have no way to know what standard they failed to meet or how to correct course. This is fundamentally inconsistent with fair contracting.
Conclusion
The federal government’s ability to access world-class AI depends on keeping commercial vendors willing and able to participate in the procurement process. This clause, as drafted, would drive many of them away to the detriment of government efficiency and affordability.
We urge the GSA to narrow the license grant to uses directly tied to contract performance, affirm that contractors retain rights in their background IP and independently developed innovations and limit transparency obligations to information vendors can realistically access and disclose without sacrificing proprietary interests. These changes would allow the government to procure AI responsibly without dismantling the commercial framework that makes that AI possible.
Sincerely,
Patrick Hedger
Director of Policy, NetChoice (The views of NetChoice expressed here do not necessarily represent the views of all NetChoice members.)
NetChoice is a trade association that works to protect free expression and promote free enterprise online.