WASHINGTON, D.C. (Federal News Network) — A lawsuit alleging that the Army misused artificial intelligence in bid evaluation before awarding a $450 million contract could shape how federal agencies document and disclose AI’s role in procurement decisions.
TRAX International Corporation is asking the U.S. Court of Federal Claims to order the Army to reevaluate the proposals submitted for the White Sands Missile Range mission support services and make a new award decision, arguing the technical evaluation was distorted by AI hallucinations that weakened TRAX’s bid.
The lawsuit comes after the Government Accountability Office denied TRAX’s bid protest challenging the Army’s contract award.
GAO acknowledged that the Army erroneously assigned TRAX a weakness but said that “this harmless error did not prejudice the protester’s chance of receiving the award.”
TRAX said the Army did not explain how the mistake occurred.
“The weakness appears to be a classic AI hallucination, with made-up references to TRAX’s proposal, that no one on the Source Selection Evaluation Board checked before sending the SSEB report to the Source Selection Authority,” the lawsuit says.
According to the Army, the Source Selection Authority did not use AI to evaluate proposals, but TRAX said the service “refused to say” whether the SSEB relied on AI during its evaluation.
“This is critical because the record that the Army produced to GAO demonstrates that the SSA simply adopted the strengths, weaknesses and other findings identified in the SSEB’s report,” TRAX said in the lawsuit.
“The record produced to GAO therefore does not explain whether the strengths assigned to [Southwest Range Services] and challenged by TRAX were identified by a member of the SSEB or the Army’s AI tool. This court, of course, refers to agency judgments and discretion, not algorithms,” the contractor added.
Legal experts say this lawsuit is notable because AI has become such a prominent element of the case, and if it survives a motion dismissed, it could shape how agencies document AI use in source selection evaluation.
Currently, there is no specific federal acquisition rule governing how agencies can use AI during bid evaluations — while procurement rules establish who has authority to make decisions, they do not address the extent to which AI can be used.
The absence of an AI-specific rule makes it harder for contractors to prove that the government violated the law, but it also does not shield agencies from legal scrutiny.
“Now agencies have to prove a human meaningfully reviewed and stood behind whatever AI produced, and if that review is not well documented, then automatically it’s going to look arbitrary, which is the standard that the contractor needs to win,” Ira Rushing, an associate at the Tully Rinckey law firm, told Federal News Network.
Regardless of the outcome of the case, Protorae Law member Alan Chvotkin said contractors are likely to start pressing agencies to disclose whether they plan to use AI during the procurement process and how it will be used.
“At least trying to pull out that disclosure. It probably won’t have much impact on the proposal response, but it might. Knowing they are using it for compliance checking is different than if they are using it for proposal evaluations,” Chvotkin told Federal News Network. “The agency will probably tell you a little bit, but not a lot.”
“Regardless of the outcome, I would hope that the takeaway here is better disclosure upfront in the solicitation of whether AI is going to be used and where. And then if it is used, what steps has the human being taken to validate the recommendations, even if by other humans who have relied on AI,” he added.
While experts say it may be difficult for TRAX to prove that the Army’s evaluation and decision to award the contract to its rival Southwest Range Services was “arbitrary and capricious,” if the case survives a motion to dismiss, the discovery process could shed light on how the Army used AI during the procurement process.
“There’s a really cool opportunity here for the contractor to not just try to win their case, but also be the first people with the opportunity to force, to see how much we can force the government to turn over AI information and discovery, because it is working everywhere else right now where courts are going, ‘If you want to use AI, you need to know that it is discoverable information.’ If they survive motion dismiss, then discovery will be where a lot of the news will come from,” Rushing said.
