GSA AI procurement rule draws fire over open-source exclusion

Industry pushback intensifies on federal AI contracting standards

The General Services Administration faces mounting criticism over a proposed acquisition rule that technology companies say could effectively bar open-source and third-party artificial intelligence from federal contracts. Despite revisions and stakeholder outreach, more than 75 comments submitted this month reveal persistent industry concern that the rule misallocates regulatory responsibility and could discourage AI adoption across government.

Major technology providers including Palantir, Nvidia, and Microsoft submitted detailed objections, focusing particularly on how the rule would handle open-source large language models and third-party AI weights. The central complaint: GSA's framework places compliance obligations on entities that lack either control over the AI systems in question or access to the government data those systems process.

Why it matters

The federal government is racing to adopt AI capabilities while managing security and bias risks. How GSA resolves these contracting questions will determine whether agencies can access cutting-edge commercial AI tools or face a narrowed vendor pool that excludes open-source innovation. The outcome could set a precedent for AI procurement across the entire federal enterprise.

Responsibility gap threatens open models

Nvidia's chief external affairs officer Bruce Andrews argued the rule creates an impossible situation for open-weight model publishers who may never interact with government data. "Our AI leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector," Andrews wrote, citing a July letter co-signed by more than 230 companies including Meta, Google, and OpenAI.

That coalition warned that excluding open models would force agencies into vendor lock-in with closed systems, which are "not inherently safe" and can fail in ways outsiders cannot detect. The White House recently indicated it would review only closed models for security risks, though that stance could evolve, according to reporting by The New York Times.

Palantir took the opposite perspective, calling it "fundamentally flawed and irrational" to hold platform providers accountable for third-party models they do not control. The company went further, formally urging GSA to withdraw the rule entirely and questioning the agency's authority to issue governmentwide procurement standards.

Practical concerns mount

The Coalition for Common Sense in Government Procurement warned that "overly broad language" could trigger a "chilling effect on contractor use and adoption of AI." Members cited impractical requirements that assume contractors own language models rather than simply operate them.

Microsoft expressed concern that the proposed terms "will significantly alter existing commercial procurement frameworks" and may require costly re-engineering that deters commercial AI sales to government. Multiple commenters suggested contractors may bypass GSA vehicles altogether if the rule stands, weakening the agency's role as a centralized technology source.

Unbiased AI definition remains contentious

GSA revised its "unbiased AI principles" definition to emphasize historical accuracy, scientific inquiry, and objectivity while maintaining neutrality. The changes did not satisfy critics. Palantir noted the clause would allocate all risk to contractors if a government official determines a commercial model fails to meet the standard.

A coalition of privacy organizations including the Center for Democracy and Technology and Electronic Frontier Foundation called for removing the unbiased principles requirement entirely, arguing it is "not technically feasible" and could be weaponized to target specific political viewpoints by any administration.

Nvidia proposed a simpler framework: "Obligations should follow the data, not the model's authorship."

The Alliance for Digital Innovation warned that if GSA cannot reconcile these requirements with standard commercial offerings, "providers may be unable or unwilling to make their most capable AI-enabled solutions available through GSA vehicles," pushing agencies toward alternative acquisition pathways that reduce competition.

These details were first reported by FedScoop.