Sanction teardown · E.D. Pennsylvania, USA · 2025-04-10
Bevins v. Colgate-Palmolive Co.
What happened
In E.D. Pennsylvania, 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)Attorney cited a non-existent Pa. Superior Court case; the WL cite maps to Dolberry v. Jakob (N.D.N.Y. 2019) and the docket number to an unrelated criminal appeal; the captioned Tinch case exists only as a 1997 N.C. decision and is irrelevant.
-
Fabricated (Case Law)Attorney cited a purported E.D. Pa. decision the Court could not locate; the WL cite maps to Top Choice Distributors v. U.S. Postal Serv. (W.D.N.Y. 1997), the case number corresponds to Daniels v. Copley Pharma., and the date does not match.
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
Striking of Counsel’s Appearance + Referral to Bar Authorities + Client Notification Order
Additional detail
AI UseCounsel filed opposition briefs citing two nonexistent cases. The court suspected generative AI use based on "hallucination" patterns but Counsel neither admitted nor explained the citations satisfactorily. Failure to comply with a standing AI order aggravated sanctions.Hallucination DetailsTwo fake cases cited. Citation numbers and Westlaw references pointed to irrelevant or unrelated cases. No affidavit or real case documents were produced when ordered.Ruling/SanctionCounsel's appearance was struck with prejudice. The Court ordered notification to the State Bar of Pennsylvania and the Eastern District Bar. Consel was required to inform his client, Bevins, of the sanctions and the need for new counsel if re-filing.
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/244/Bevins_v._Colgate_Palmolive_US_DC_Eastern_Penn._April_10_2025.pdf, via Damien Charlotin's public AI Hallucination Cases Database (CC0).