What was hallucinated

Fabricated: Case Law | Defense cited a non-existent case to argue privilege assignments must be clear and unmistakable; Plaintiff could not find it on Westlaw/Lexis, the Court also could not locate it, and defense later withdrew and admitted AI use.

Details

AI UseFirst Counsel, who had not previously used AI for legal work, used an unspecified AI tool to assist with drafting a response. He failed to verify the citation before submission. Second Counsel, as local counsel, filed the response without checking the content or accuracy, even though he signed the document.Second Counsel then said that he had initiated "procedural safeguards to prevent this error from happening again by ensuring he, and local counsel, undertake a comprehensive review of all citations and arguments filed with this and every court prior to submission to ensure their provenance can be traced to professional non-AI sources."Hallucination DetailsThe hallucinated case was cited as controlling Delaware authority on privilege assignments. When challenged by Plaintiff, Defendants initially filed a bare withdrawal without explanation. Only upon court order did they disclose the AI origin and acknowledge the error. Counsel personally apologized to the court and opposing counsel.Ruling/SanctionJudge William Matthewman imposed a multi-part sanction:Attorneys’ fees and costs incurred by Plaintiff in rebutting the hallucinated citation—jointly payable by CounselRequired CLE on AI ethics within 30 days, with proof of completion due by June 20, 2025Monetary fines: $1,000 (First Counsel) and $500 (Second Counsel), payable to the Court registryThe Court emphasized that the submission of hallucinated citations—particularly when filed and signed by two attorneys—constitutes reckless disregard for procedural and ethical obligations. Though no bad faith was found, the conduct was sanctionable under Rule 11, § 1927, the Court’s inherent authority, and local professional responsibility rules.Key Judicial ReasoningThe Court distinguished this case from more egregious incidents (O’Brien v. Flick, Thomas v. Pangburn) because the attorneys admitted their error and did not lie or attempt to cover it up. However, the delay in correction and failure to check the citation in the first place were serious enough to warrant monetary penalties and educational obligations.

Sanction teardown · S.D. Florida, USA · 2025-05-20

Versant Funding v. Teras Breakbulk Ocean Navigation Enterprises

What happened

In S.D. Florida, 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)

Defense cited a non-existent case to argue privilege assignments must be clear and unmistakable; Plaintiff could not find it on Westlaw/Lexis, the Court also could not locate it, and defense later withdrew and admitted AI use.

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

Joint and several liability for Plaintiff’s attorneys' fees and costs incurred in addressing the hallucinated citation; CLE requirement on AI ethics; Monetary fines (monetary penalty: 1500 USD.)

Additional detail

AI UseFirst Counsel, who had not previously used AI for legal work, used an unspecified AI tool to assist with drafting a response. He failed to verify the citation before submission. Second Counsel, as local counsel, filed the response without checking the content or accuracy, even though he signed the document.Second Counsel then said that he had initiated "procedural safeguards to prevent this error from happening again by ensuring he, and local counsel, undertake a comprehensive review of all citations and arguments filed with this and every court prior to submission to ensure their provenance can be traced to professional non-AI sources."Hallucination DetailsThe hallucinated case was cited as controlling Delaware authority on privilege assignments. When challenged by Plaintiff, Defendants initially filed a bare withdrawal without explanation. Only upon court order did they disclose the AI origin and acknowledge the error. Counsel personally apologized to the court and opposing counsel.Ruling/SanctionJudge William Matthewman imposed a multi-part sanction:Attorneys’ fees and costs incurred by Plaintiff in rebutting the hallucinated citation—jointly payable by CounselRequired CLE on AI ethics within 30 days, with proof of completion due by June 20, 2025Monetary fines: $1,000 (First Counsel) and $500 (Second Counsel), payable to the Court registryThe Court emphasized that the submission of hallucinated citations—particularly when filed and signed by two attorneys—constitutes reckless disregard for procedural and ethical obligations. Though no bad faith was found, the conduct was sanctionable under Rule 11, § 1927, the Court’s inherent authority, and local professional responsibility rules.Key Judicial ReasoningThe Court distinguished this case from more egregious incidents (O’Brien v. Flick, Thomas v. Pangburn) because the attorneys admitted their error and did not lie or attempt to cover it up. However, the delay in correction and failure to check the citation in the first place were serious enough to warrant monetary penalties and educational obligations.

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/373/Versan_Funding_v._Teras_Breakbulk_S.D._Florida_USA_May_20_2025.pdf, via Damien Charlotin's public AI Hallucination Cases Database (CC0).