Palantir calls on GSA to withdraw draft AI acquisition rule

The General Services Administration should axe its draft large language model-specific acquisition rule, in part because it is purportedly unlawful and would undermine Trump administration goals if it were finalized, Palantir argued in recent comments on the proposal.

“If implemented, the Proposed AI Clause will likely prevent Palantir from providing government customers with the AI-powered solutions they need through GSA contracts,” law firm Freshfields warned on behalf of the company. 

“Accordingly, Palantir expects government customers to turn to non-GSA contract vehicles to acquire Palantir’s most valuable AI-powered commercial solutions, thus diluting the promise of GSA contracts overall,” the company added in a letter to GSA.

Palantir argued that GSA has no independent authority to finalize the proposed LLM rule and would violate the Federal Acquisition Streamlining Act by mandating contracting requirements beyond standard commercial practices if it did finalize it. It notes that the courts have held that when an agency claims authority to decide on an issue of “vast economic and political significance,” Congress must delegate that authority in statute, which it has not done here. The company further said that the rule would significantly burden contractors and cause agencies to turn to other contract vehicles, undermining the Trump administration’s goal of reducing inefficiencies and consolidating procurement through GSA.

The company also called the proposed LLM rule incompatible with legal requirements for defense acquisitions — such as limitations on the clauses defense agencies can require of contractors and prohibitions on “demanding proprietary IP rights as a condition of contract award” — and said it disregards decades of intellectual property rights precedence.

Menaka Kalaskar, head of Palantir’s U.S. government legal and contracting team, raised much of the same concerns during a GSA listening session last month but stopped short of calling for the agency to withdraw the rule.

“If the clause is really needed, then it can be included in the [Federal Acquisition Regulation] overhaul rather than the GSA coming out on its own,” Kalaskar said at the time.

Palantir’s call for GSA to kill the LLM acquisition clause stands apart from much of the rest of the federal contracting industry. 

During last month’s listening session, industry groups representing contractors widely urged GSA to clarify definitions and modify flowdown requirements if the agency wanted to achieve its goal of safeguarding government data processed by LLMs. Groups reiterated many of those same talking points in letters sent to GSA, according to an Nextgov/FCW review of a dozen public comments on regulations.gov.

Coalition for Common Sense in Government Procurement members “continue to have concerns regarding the overly broad language utilized in the AI clause and believe the AI clause may have a chilling effect on contractor use and adoption of AI should the clause be finalized as written,” the group wrote to GSA. 

The coalition asked GSA to ensure only data directly submitted to an LLM’s inference engine — not upstream and downstream processing by ancillary system components — be subject to LLM rule requirements. The group also asked GSA to create a bright-line test to clearly define when LLM functionality is “incidental to the primary purpose of the core requirement being procured” and therefore not subject to LLM rule requirements.

The Software & Information Industry Association, AEM Corporation, Sheppard, Mullin, Richter & Hampton LLP, the Business Software Alliance, Professional Services Council and the Information Technology Industry Council similarly called for GSA to more clearly define the scope of the LLM acquisition rule, Nextgov/FCW found in a review of comments sent to GSA.

But the contracting industry is not united on how to reform the proposal.

For example, the Information Technology Industry Council urged GSA to get rid of a requirement that contractors ensure LLMs are developed and monitored in a way that adheres to “unbiased AI principles.”

“The criteria through which the government will evaluate compliance are still insufficiently defined, which creates ongoing uncertainty regarding the government’s expectations,” Megan Petersen, ITI’s senior vice president of policy, public sector and counsel, told Nextgov/FCW. “Without clear definitions and transparent compliance benchmarks, contractors will be held to an impossible performance standard.”

Other groups instead urged GSA to clear up the “unbiased AI principles” language with objective metrics contractors could use to ensure compliance with the regulation.

The Professional Services Council urged GSA to publish its benchmark methodology and evaluation criteria for “unbiased AI principles” and update them through notice-and-comment periods. The group pointed to NIST AI RMF 1.0 trustworthiness criteria and the IEEE 7003 Algorithmic Bias standard as some third-party standards GSA could use.

In total, 79 stakeholders sent GSA comments on the potential LLM procurement overhaul before the Aug. 3 deadline, according to regulations.gov.