Sanction teardown · Supreme Court of The Bahamas, The Bahamas · 2025-08-01
Marcus Groesser and Ira Hess v. Robert Phelps Herman
What happened
In Supreme Court of The Bahamas, The Bahamas, a filing relied on ChatGPT to help draft legal argument. The court identified the following problems with the citations in that filing:
-
Fabricated (Case Law)Non-existent Kelly v Rolle cited in oral submissions; court confirmed the case does not exist and the citation led to an unrelated decision (Hall v The Attorney General).
-
Fabricated (Case Law)Non-existent Petrie v Dowling cited; court confirmed no such [1992] 1 WLR 1017 case exists.
-
Fabricated (Case Law)Non-existent Ladmat Ltd v Backo cited; court found the citation led to CVC v Almeida and that the asserted holding was fabricated.
-
Fabricated (Case Law)Later submission of a fabricated appellate citation for Kelly v Rolle; court noted impossible panel date (post-retirement of Conteh JA) and mismatched file number.
Which AI tool
ChatGPT. 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
Overall issue of hallucination referred to the bar; Order to bear costs (monetary penalty: 1 .)
Additional detail
"37. The Court does not accept that there is a difference between the oral and written submissions. In fact, had there not been a request by the Plaintiff to respond, the Court would have ruled on the oral submission and such ruling could have significantly relied on the submissions of Defence Counsel. The implications are severe and serious when the Court cannot accept Counsel's assertion to be truthful and cases to be real. The risk of harm to the integrity of the judicial process is real and could bring the system into disrepute. [...]40. The Court does not accept Ms. Taylor's submission that the fictitious cases were verified before layover and that they were only used in speaking points and were "not intended to form the official record." I find there is no distinction between the speaking points, oral submissions and written submission. They are all submissions advanced by Counsel intended for the Court to rely on them in the process of decision making. The purpose of which was to guide a judgment in your client's favour. The attempt to draw such a distinction is one without merit and the Court rejects same without more."
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/658/Gruesser-Judgment.pdf, via Damien Charlotin's public AI Hallucination Cases Database (CC0).