Government Contracting

On the last day of March 2025, the Court of Federal Claims (COFC) issued the first procurement decision openly discussing “GenAI Misuse.”^[1] I published a comprehensive 2025 report on GenAI Misuse in procurement litigation at the beginning of 2026.^[2] There, I defined GenAI Misuse as the intentional or negligent use of an LLM resulting in inaccuracies that waste public and private resources.

Over halfway through 2026, GenAI Misuse is increasing, particularly among lawyers. Research shows that this increase also does not account for a significant amount of non-public GenAI Misuse. One consolation is that while GenAI Misuse increases, it still almost always sabotages chances for success in litigation. I continue to advocate that tribunals should take a more aggressive and public stance on GenAI Misuse, following the lead of the Armed Services Board of Contract Appeals (ASBCA).

First, GenAI Misuse is increasing in 2026. Prior to 2025, there were no public final decisions at any of the procurement tribunals mentioning GenAI Misuse. In 2025, there were twenty-three cases of GenAI Misuse and three sanctions.^[3] In 2026, there are already at least 20 cases involving GenAI Misuse as of early August, including three where a tribunal has issued sanctions.

The current tally is an undercount because many instances of GenAI Misuse are never made public. Public information requests have helped unveil additional and previously unknown GenAI Misuse in filings when the tribunal does not mention GenAI Misuse. In some cases, I identified clear case law hallucinations in filings, even though there is no mention of the misuse by the tribunal or any party.^[7] These discoveries usually rely on issuing a public information request on a hunch. Public information requests are also fulfilled slowly, and many filings are covered by protective orders, including extensive redactions or preventing disclosure altogether.

The bigger issue is that most procurement cases never reach a final decision—they settle, are dismissed, or are withdrawn before a final resolution. Indeed, this problem is more acute because GenAI Misuse is co-morbid with unsuccessful bid protests and claims. In other words, GenAI-related errors make it more likely that a case ends prematurely, meaning it is also more likely that there will be no public decision discussing the misuse.

A related problem is that tribunals have mostly identified GenAI Misuse as citations to fake cases. That is because it is simple to see whether a case exists or not. It is much harder to identify other forms of GenAI Misuse, like when facts and quotations from real cases are misquoted or misrepresented. This type of misuse is more likely to go unremarked upon because it is the type of error a human might also make and is not necessarily indicative of GenAI Misuse.

For instance, in a February 2026 CBCA decision, the judges noted that a pro se litigant’s motion contained “four excerpts from depositions in quotation marks. . .” where the “quoted language did not exist on the pages cited or anywhere in the deposition transcripts.” ^[8] Tribunals seem to be recognizing that LLM factual errors are more widespread than fabricated authority. Other cases defy neat categorization, such as sycophancy, which occurs when a GenAI platform provides an output that is designed to agree with the position taken by the GenAI user, regardless of the veracity of that position.^[9]

Second, and more troubling, lawyers are now a substantial part of the problem. In 2025, the first law firm GenAI Misuse decision came from the ASBCA, which struck a brief after finding that over 70 percent of the citations were inaccurate.^[10] The ASBCA wrote: “Our power to impose sanctions is broad and may even extend to dismissal of an appeal.” The ASBCA has issued four 2026 decisions condemning GenAI Misuse, compared to two decisions from 2025.

In one protest, the lawyers admitted to using Westlaw CoCounsel, a legal-specific GenAI tool.^[11] Counsel said they identified and corrected some “irregularities” before filing, but “admittedly failed to correct” all of them. In another bid protest decision, GAO chastised a lawyer after they failed to “address the basis for the inaccurate case citations, as we expressly requested.”^[12] In other cases, discussion of the GenAI Misuse was relegated to a footnote warning that future submissions might “result in the imposition of sanctions.”^[13]   

Third, as GenAI Misuse grows, one small consolation is that such misuse has not been rewarded. GenAI Misuse appears to seriously damage credibility such that few, if any, cases involving GenAI Misuse succeeded in obtaining bid protest relief or monetary damages. The sole exception is in a non-dispositive summary judgment briefing at the ASBCA.^[14] There, the ASBCA both sanctioned the pro se contractor for citing numerous hallucinated cases and granted the contractor’s motion for summary judgment.

Most decisions involving GenAI Misuse are on the merits, and the damage to credibility from GenAI Misuse appears to sabotage contractors’ chances of success. So far, only one case, a protest, was dismissed directly as a sanction.^[15] Between April 2025 and July 2026, the same contractor filed at least 23 protests at GAO, including 11 that were filed after GAO sanctioned the contractor for repeated GenAI Misuse in September 2025. Every protest resolved so far, even those without clear signs of GenAI Misuse, has been dismissed or denied, taking an average of 48 days to resolve. In addition to the twenty-three GAO bid protests, the same contractor has filed at least two monetary claims at the Civilian Board of Contract Appeals (CBCA). This contractor appears to have been unsuccessful in all twenty-five of these cases.

There are a few takeaways. GenAI Misuse is a serious and growing problem for procurement tribunals. 2026 has been significantly more troubling than 2025. We will see if the trend continues through the end of the year. In the meantime, tribunals can take concrete steps to mitigate the greater ills attendant to GenAI Misuse.

I wrote about the threat cynical pro se protesters using GenAI pose to the overall procurement system for National Contract Management Association (NCMA) earlier this year. GAO may be able to deal with a few of these serial filers, but what happens when more contractors adopt the same strategy? What tools should the tribunals and/or the Government be reaching for to discourage this behavior?

The solution need not be complicated. Procurement tribunals should require a pre-filing disclosure and certification that facts and citations have been verified. This solution appears to have been partially adopted by the ASBCA, which features a prominent tab on its website warning filers that “the current generation of AI tools are known to sometimes create materially false characterizations of legal precedent, misquote cases, and even create non-existent case citations.”

The benefit of a pre-filing warning is to warn pro se and non-technologically savvy lawyers that GenAI might create plausible but false statements and citations. Those reading my work are very familiar (perhaps even bored) with this issue. Nevertheless, there appear to be many who do not know of the problem. Likewise, the warning serves to sharpen the ability of the procurement tribunals to sanction and manage their docket to mitigate rampant misuse.

Consequences for lawyers should probably be more severe than for pro se litigants, given their professional ethical obligations. GAO has already warned that fake citations might result in referral to appropriate bar associations or other disciplinary bodies. It is unknown whether GAO or any other procurement tribunal has actually referred a lawyer for discipline related to GenAI Misuse. Tribunals should create an internal policy to alert practitioners to the potential consequences.

More research into this issue is necessary to uncover the extent to which GenAI Misuse is impacting the procurement system. The tribunals are probably best positioned to adopt policies to stem the tide of misuse and to publish research based on their access to the whole universe of problems. These research efforts are likely to need special funding, which might require new legislation. I hope to shed light on the issue to spur this effort forward through my research.

Footnotes

[1] Sanders v. United States, 176 Fed. Cl. 163 (2025).

[2] The procurement tribunals include the Court of Federal Claims, the Government Accountability Office, the Armed Services and Civilian Board of Contract Appeals.

[3] My research has uncovered at least three previously unknown instances, not present in my original report, increasing the 2025 total to twenty-three.

[4] Current through the end of July, 2026.

[5] A “Decision” means a final or interim disposition of a case or filing in a written finding by the tribunal.

[6] A “Sanction” is a specific exercise by the tribunal of its inherent authority to punish the contractor for Gen-AI Misuse.

[7] See The JAAW Group, LLC, B-424133 (February 2, 2026) (failing to mention GenAI Misuse in decision); Seed v. United States, No. 23-cv-01947 (June 9, 2026) (same).

[8] Blazy v. Department of State, CBCA Nos. 7992, 7993 (Feb. 24, 2026).

[9] Appeal of Sayar Development, Inc., ASBCA No. 63871 (March 17, 2026) (noting “editorial comments provided by artificial intelligence. . .” in appellant’s briefing).

[10] Appeal of Huffman Constr., LLC, ASBCA Nos. 62591, 62783 (Oct. 23, 2025) (striking a reply brief as a sanction for GenAI Misuse).

[11] CVTEK, LLC, B-423943; B-423943.2 (Feb. 12, 2026).

[12] Bramstedt Surgical Inc., B-424064 (January 28, 2026).

[13] G2 Ops, Inc., B-423967 (February 13, 2026); Adapt Forward, LLC, B-424006 (February 18, 2026).

[14] PDO Ahuska Joint Venture LLC, ASBCA No. 64164 (July 13, 2026).

[15] Oready, LLC, B-423649.1; B-423650.1; B-423670.1; B-423670.2 (Sep. 25, 2025) (dismissing four protests due to repeated GenAI Misuse).

David P.J. Timm, Procurement is Not "Oready" for GenAI Misuse, Contract Management (January–February 2026)