What was hallucinated

Fabricated: Exhibits & Submissions | Application to reopen attributed an inflammatory prosecutor quotation cited to the trial transcript p.559; court found p.559 was the court reporter's signature page and no such quotation existed — the quote was generated by ChatGPT and incorporated into the filing without verification. || Fabricated: Exhibits & Submissions | Application cited additional quotations on pages 2 and 4 of appellant's brief (including an alleged 'duty to retreat' reference) that could not be located in the record; court determined these were wholesale inventions produced by ChatGPT and that they impugned the prosecutor, judge, and prior counsel.

Details

"{¶133} This case illustrates the peril. An attorney who, by his own counsel’s admission, was sophisticated in his understanding of AI tools permitted a non-attorney staff member to use a public generative AI platform to prepare an appellate filing. The AI tool fabricated transcript quotations—attributing specific, inflammatory statements to a real prosecutor that were never spoken. The attorney filed the document without verifying its contents. When the fabrications were identified, he did not correct the record. He appealed this court’s denial of the tainted application to the Supreme Court of Ohio without disclosing the fabrications. He proffered an AI policy that itself appeared to have been generated by AI, complete with unfilled placeholder brackets. Two months after a sanctions hearing, a filing in another court bore the unmistakable hallmarks of unchecked AI output, including a ChatGPT prompt embedded in the text of a legal brief.{¶134} This court does not write to condemn the use of artificial intelligence in the practice of law. To the contrary, this court recognizes that AI is an inevitable and potentially beneficial feature of modern legal practice. But the use of AI does not relieve an attorney of any of the obligations imposed by the Rules of Professional Conduct, by the rules of court, or by the oath of admission to the bar. An attorney who files a document containing AI-generated content is responsible for that content, fully and without qualification. The duty to verify, the duty of candor, the duty of competence, and the duty of supervision cannot be delegated to a machine.{¶135} The sanctions imposed herein are proportionate, individually justified, and collectively designed to serve the purposes for which the court’s sanctioning authority exists: to compensate for harm, to deter future misconduct, to protect the integrity of the judicial process, to preserve public confidence in the administration of justice, and to ensure that the practice of law remains a profession grounded in truth, accuracy, and candor. "

Sanction teardown · CA Ohio, USA · 2026-03-20

State v. Coleman

What happened

In CA Ohio, USA, a filing relied on ChatGPT to help draft legal argument. The court identified the following problems with the citations in that filing:

Fabricated (Exhibits & Submissions)

Application to reopen attributed an inflammatory prosecutor quotation cited to the trial transcript p.559; court found p.559 was the court reporter's signature page and no such quotation existed — the quote was generated by ChatGPT and incorporated into the filing without verification.
Fabricated (Exhibits & Submissions)

Application cited additional quotations on pages 2 and 4 of appellant's brief (including an alleged 'duty to retreat' reference) that could not be located in the record; court determined these were wholesale inventions produced by ChatGPT and that they impugned the prosecutor, judge, and prior counsel.

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

Monetary Sanction; Bar Referral; Counsel disqualified in this case; Order to notify the judgment in other cases; CLE; Letter of apology (monetary penalty: 2000 USD.)

Additional detail

"{¶133} This case illustrates the peril. An attorney who, by his own counsel’s admission, was sophisticated in his understanding of AI tools permitted a non-attorney staff member to use a public generative AI platform to prepare an appellate filing. The AI tool fabricated transcript quotations—attributing specific, inflammatory statements to a real prosecutor that were never spoken. The attorney filed the document without verifying its contents. When the fabrications were identified, he did not correct the record. He appealed this court’s denial of the tainted application to the Supreme Court of Ohio without disclosing the fabrications. He proffered an AI policy that itself appeared to have been generated by AI, complete with unfilled placeholder brackets. Two months after a sanctions hearing, a filing in another court bore the unmistakable hallmarks of unchecked AI output, including a ChatGPT prompt embedded in the text of a legal brief.{¶134} This court does not write to condemn the use of artificial intelligence in the practice of law. To the contrary, this court recognizes that AI is an inevitable and potentially beneficial feature of modern legal practice. But the use of AI does not relieve an attorney of any of the obligations imposed by the Rules of Professional Conduct, by the rules of court, or by the oath of admission to the bar. An attorney who files a document containing AI-generated content is responsible for that content, fully and without qualification. The duty to verify, the duty of candor, the duty of competence, and the duty of supervision cannot be delegated to a machine.{¶135} The sanctions imposed herein are proportionate, individually justified, and collectively designed to serve the purposes for which the court’s sanctioning authority exists: to compensate for harm, to deter future misconduct, to protect the integrity of the judicial process, to preserve public confidence in the administration of justice, and to ensure that the practice of law remains a profession grounded in truth, accuracy, and candor. "

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/1774/2026-Ohio-965.pdf, via Damien Charlotin's public AI Hallucination Cases Database (CC0).