Sanction teardown · GAO, USA · 2025-08-13

Oready, LLC v. GAO

What happened

In GAO, USA, a filing relied on an unnamed/unconfirmed AI tool to help draft legal argument. The court identified the following problems with the citations in that filing:

• Fabricated (Case Law)
  Protester cited 'Sayres & Assocs. Corp., B-418374.3, July 20, 2020, 2020 CPD ¶ 242,' but GAO stated no such decision exists; the only Sayres decision (B-418374, Mar. 30, 2020) is unrelated and does not support the proposition.
• Fabricated (Case Law)
  Protester cited 'Bannum Inc., B-416107.2, June 12, 2018, 2018 CPD ¶ 204,' but GAO found no such B-number exists; none of Bannum’s GAO decisions support the asserted proposition.
• Misrepresented (Case Law)
  Protester cited 'Total Health Res., B-414101, Feb. 21, 2017, 2017 CPD ¶ 64,' but GAO noted the accurate decision is Total Health Resources, B-403209, Oct. 4, 2010, 2010 CPD ¶ 226; thus the reference was inaccurate.
• Fabricated (Case Law)
  Protester cited 'BluePath Labs, LLC, B-421791, Aug. 4, 2023,' but GAO found B-421791 is an unpublished Pitney Bowes decision; the BluePath Labs citation does not exist and is irrelevant.
• Fabricated (Case Law)
  Protester cited 'GTA Containers, Inc., B-411556, Sept. 16, 2015, 2015 CPD ¶ 281,' but GAO explained B-411556 is an unpublished SSI Tech. Inc. decision; the cited GTA Containers decision does not exist and is irrelevant.

Which AI tool

an unnamed/unconfirmed AI tool. Note: Charlotin's public database records tool attribution only where a court order, brief, or reporting on the matter states it explicitly; "unidentified" or "implied" means the record indicates AI use but does not name a specific product — we do not guess.

Outcome

Warning

Additional detail

"Second, the protester's explanation--that it was “manual mismatches in secondary summaries” that caused the citation errors (Protester's Resp., Aug. 8, 2025, at 1)--does not meaningfully explain the number of citation errors in the protester's filings. Indeed, Oready's patently erroneous citations are far removed from mere typographical or scrivener's errors, and instead, bear the hallmarks of the use of a large-language model or other artificial intelligence (AI) without adequate verification that the generated results were accurate. "

How Citation Safe would have caught this

Citation Safe runs three deterministic layers before a brief is filed: (1) does the citation exist against CourtListener's database of published opinions, (2) if quoted, does that exact language appear in the source, (3) does the cited case actually support the proposition it is cited for. Fabricated case citations fail Layer 1. Fabricated or misattributed quotations fail Layer 2 even when the underlying case is real. Misrepresented holdings — a real case cited for a proposition it does not support — are the target of Layer 3. None of these checks involve asking another language model whether the citation looks right; they are lookups and text-matches against the actual source, which is why a hallucinated citation has to survive a direct lookup against the authoritative source — not another model's opinion — to earn a VERIFIED stamp; our measured false-verify rate is published live at /quality.

Check a brief before you file it → · See our live false-verify rate

Source: https://www.damiencharlotin.com/documents/663/Matter_of_Oready_USA_13_August_2025.pdf, via Damien Charlotin's public AI Hallucination Cases Database (CC0).