Sanction teardown · District Court, China · 2026-09-15
Credit One Finance Limited v Cheung Yiu Cho and another
What happened
In District Court, China, 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)D2 cited a nonexistent or untraceable Hong Kong case. The Court stated that it could not locate the authority and rejected the alleged citation.
-
Fabricated (Case Law)D2 cited a nonexistent or untraceable Hong Kong case. The Court stated that it could not locate the authority.
-
Fabricated (Case Law)D2 cited a nonexistent or untraceable Court of Final Appeal case. The Court stated that it could not locate the authority.
-
False Quotes (Case Law)D2 attributed paragraphs 51 and 52 to the case, but the judgment contained only 39 paragraphs. The Court also found the case irrelevant to the appeal.
-
False Quotes (Case Law)D2 cited Ling Wing Fai Billy as reported at (2011) 15 HKCFAR 373 and referred to paragraph 58, but the Court stated that the correct case number was FACV No. 11 of 2011, the case was reported within (2012) 15 HKCFAR 344, and the judgment contained only 39 paragraphs.
-
Misrepresented (Case Law)D2 cited CIBC Mortgages plc v Pitt with the incorrect citation 1 AC 2002. The Court corrected the citation to [1994] 1 AC 200 and found the case factually irrelevant to this dispute.
-
Misrepresented (Case Law)D2 cited Kingsnorth Trust Ltd v Tizard with an incorrect title and citation. The Court stated that the correct authority was Kingsnorth Finance Co. Ltd v Tizard, [1986] 1 WLR 783, and explained that its facts and principle did not assist D2.
-
False Quotes (Case Law)D2 cited Southern Pacific Mortgages Ltd v Scott with the incorrect citation AC 769. The Court stated that the correct citation was [2015] AC 385 and that the case did not contain the proposition attributed to it.
-
Misrepresented (Doctrinal Work)D2 attributed findings to the Court that the property was beneficially owned by D2, that D2 had paid all mortgage instalments, and that D2 was honest. The Court stated that its judgment made no such findings and instead found that D2 failed to prove those matters.
-
Misrepresented (Legal Norm)D2 asserted that the Court had held that private ownership or trust arrangements were invalid under section 17B of the Housing Ordinance. The Court stated that its earlier judgment never mentioned section 17B or made that ruling.
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
Not specified in source record.
Additional detail
The Court found that D2's written submissions were probably largely AI-generated, consistent with D2's prior admission that much of his trial advocacy had been generated by AI. The submissions contained nonexistent or untraceable cases, incorrect citations and paragraph references, and propositions that the cited authorities did not support. The Court dismissed the application for leave to appeal and the stay applications, finding no reasonable prospect of success, and ordered D2 to pay costs. The Court also rejected reliance on 111 pages of new documents because the requirements for admitting new evidence on appeal were not met.
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/3160/DCMP002540A_2021.pdf, via Damien Charlotin's public AI Hallucination Cases Database (CC0).