From the TIKC Advisory Team · AI & Technology in GovCon
AI Hallucinated and Cost a Contractor a $450M Army Award — What Every Federal Bidder Must Know Now
Trax International is suing the Army at the Court of Federal Claims, alleging AI hallucinated weaknesses in its proposal that cost it a $450M contract…

Published August 11, 2026 | TIKC NewsWire
A federal contractor is asking the U.S. Court of Federal Claims to throw out a $450 million Army contract award after alleging that the Army’s artificial intelligence tool hallucinated weaknesses in its proposal — weaknesses that never existed, fabricated by an AI system that no human reviewer caught before it influenced the award decision. The case, Trax International Corporation versus the United States, is the first major public window into how AI is being used in federal source selection — and what happens when it goes wrong.
What Happened at White Sands
The Army used AI that hallucinated multiple times, creating the false impression that Trax’s bid was weaker than rival Southwest Range Services’ successful bid of roughly $450 million, a complaint filed in late July alleged. After losing, Trax filed a bid protest at GAO — which acknowledged the Army had erroneously assigned Trax a weakness but ruled the error was harmless. Trax escalated to the Court of Federal Claims.
“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.
The Disclosure Problem
An industry lawyer anticipates more bid protests will allege that agencies misused AI in the procurement process, but that could be hard to prove if agencies don’t report their use of AI as “high-impact.” Few civilian agencies are reporting the use of AI in bid evaluations as “high-impact,” meaning agencies have either not implemented minimum risk practices or have not publicly disclosed them.
What Every Federal Contractor Must Do Right Now
Request your debrief and read it for AI hallucinations. Read every weakness citation carefully. Does the weakness reference something your proposal actually said? Are the page numbers cited real? Hallucinated weaknesses leave traces — references to things that aren’t in your proposal, misattributions, and internally inconsistent evaluator narratives.
Ask directly whether AI was used in your evaluation. In your debrief, ask specifically whether any AI tools were used in reviewing or evaluating proposals. Document the question and the response.
Preserve your protest rights. The GAO protest deadline is 10 days from the conclusion of your debrief. If you see signs of AI hallucination in your evaluation record, engage protest counsel immediately — the timeline is unforgiving.
The Bottom Line
AI in federal procurement is no longer hypothetical — it is in source selection evaluation boards right now, reading your proposals and assigning strengths and weaknesses. The Trax case proves that AI can hallucinate in this context, that those hallucinations can survive without human review, and that they can affect billion-dollar award decisions. Read your debriefs like a lawyer. Brick by brick — know what you submitted, and know what the evaluator actually read.
Frequently Asked Questions
What is the Trax Army AI bid protest case?
Trax International Corporation filed a complaint at the U.S. Court of Federal Claims alleging the Army used AI that hallucinated weaknesses in its $420 million proposal for mission support services at White Sands Missile Range. GAO had previously denied the protest but acknowledged the Army erroneously assigned Trax a weakness.
Can AI legally be used in federal proposal evaluations?
Yes. DoD has disclosed AI use in several recent solicitations for bid evaluation assistance. OMB rules require agencies to classify high-impact AI uses and apply minimum risk practices — but few civilian agencies are reporting bid evaluation AI as high-impact, creating a significant governance gap.
What is an AI hallucination in the context of a bid evaluation?
An AI hallucination is when an AI system generates information that is factually incorrect or fabricated — in this context, citing weaknesses in a proposal that do not actually appear in the proposal document. The Trax complaint alleges the Army AI cited weaknesses with references to content that did not exist in Trax’s submitted proposal.
References
Defense One. (2026, August). Did AI blow a $450M Army contract decision? Company lawsuit says yes. https://www.defenseone.com/policy/2026/08/ai-army-contract-lawsuit/415261/
Federal News Network. (2026, August 11). Lawsuit over Army’s use of AI in contract award could increase transparency around proposal evaluations. https://federalnewsnetwork.com/contracting/2026/08/lawsuit-over-armys-use-of-ai-in-contract-award-could-increase-transparency-around-proposal-evaluations/
About the Author
Melanie Patterson
Founder & CEO of Team Integrity Knowledge Center and creator of GovCon iSource. Former nurse turned entrepreneur with over 10 years guiding small, women-owned, and minority-owned businesses to over $10 million in government awards. Build, grow, scale — brick by brick. Contact
Your next step
Talk it through with a TIKC advisor
Continue reading
Related stories

DHS Issues $2M Procurement Notice for Boston Dynamics Robot Dogs — What It Signals for the Federal Robotics Market
ICE just dropped a DHS Acquisition Planning Forecast notice for up to $2 million in Boston Dynamics Spot robots for hazardous environment inspection and law…

CISA Eyes a Follow-On Cybersecurity Operations Support Contract — How Small Cyber Firms Can Get In Now
CISA is developing a follow-on to its cybersecurity operations support contract — one of the most valuable recurring vehicles in the federal cyber market. Here…

Peraton Wins $953M DISA Recompete — What the Award Means for Small Business IT Contractors in the Comms Infrastructure Market
Peraton secured the $953M DISA communications infrastructure recompete — one of the most significant DISA awards of the year. Here is what it means for…

The $1 Federal AI Deals Are Expiring September 30 — 3.4 Million Government Workers Face an AI Cliff
GSA’s OneGov deals giving 3.4 million federal workers access to ChatGPT, Gemini, and Claude for $1 expire September 30 with no public successor plan. The…

The Biggest Veteran-Care Reorganization in 30 Years Is Awarding Now: Where Small Firms Fit in VA CCN Next Gen
VA’s CCN Next Gen — a decade-long, hundreds-of-billions community care overhaul — begins awarding in August. The primes are health plans; the opportunity for small…

The $1 AI Deal Reshaping Federal Buying: What GSA’s OneGov Means for Your Business
GSA’s OneGov strategy is putting frontier AI in agencies’ hands for pennies. Here’s what the shift to centralized commercial buying means for small GovCon firms.

The Pentagon’s Fastest Money Just Opened a Software Lane: APFIT Tops $2 Billion
DoD’s APFIT program passed $2B and made its first software-only awards. Built for small and non-traditional firms, it’s one of the fastest paths into defense.

The SBIR Reset: New Proposal Caps and ‘Strategic Breakthrough’ Awards Land in FY2027
SBIR/STTR was reauthorized through 2031 — but the reforms bite in FY2027: agency proposal caps, a new $30M Strategic Breakthrough award, and tighter foreign screening.

GSA’s New AI Clause Lands August 3: What LLM Contractors Must Know Before the Comment Window Closes
GSA’s revised LLM data-safeguarding clause 552.239-7001 imposes new obligations on AI contractors. Comments close August 3, 2026. What to know now.

Quantum Just Got a $125M Federal Vote of Confidence — And a Chicago Footprint. Here’s the GovCon Angle.
PsiQuantum signed a $125M DARPA agreement under the Quantum Benchmarking Initiative. What the federal quantum push means for small contractors — and Chicago.

Quantum Just Became a Procurement Problem: New Federal Mandate Sets 2030 and 2031 Deadlines
EO 14412 sets 2030 and 2031 post-quantum deadlines and pushes new FAR rules for covered contractors. What small federal suppliers should do now.




