AI Sanctions Register · Citation Hallucination in Court Filings
1The Heppner ruling concerns privilege. This register concerns a separate failure mode: attorneys who submit AI-generated citations to cases that do not exist. The two are distinct. Heppner addresses what happens to confidential information typed into a consumer AI. This page addresses what happens when an attorney files AI-generated authority without verifying it.
2The Damien Charlotin AI Hallucination Cases Database tracked 1,453 such cases worldwide as of May 2026. US courts imposed more than $145,000 in sanctions in the first quarter of 2026 alone. The cases below are documented with court, date, outcome, and primary source. Last updated June 4, 2026.
The Pattern
3A generative AI system produces text by predicting likely word sequences. When asked for a citation, it generates text that resembles a citation. If no matching case exists, it generates a plausible one anyway. This is not a malfunction. It is how the technology works. A system that generates citations cannot guarantee they correspond to real documents. Only a system that retrieves citations from an indexed corpus can make that guarantee.
4Courts have applied existing rules to this conduct. Federal Rule of Civil Procedure 11 requires that filed documents have evidentiary support. A nonexistent case fails that standard. The sanctions below escalate from monetary fines to license suspension. The trend is toward severity.
Documented Incidents
Court
U.S. Court of Appeals, Ninth Circuit
What Happened
Immigration attorneys Mike Singh Sethi and William Rounds filed briefs containing multiple nonexistent cases, misattributed quotations, and gross misrepresentations of real cases. The brief was drafted by an unlicensed writer. No licensed attorney read the cited cases before filing. The attorneys denied AI use through oral argument, a motion to correct, and a show-cause response.
Outcome
$2,500 sanction each. Six-month suspension from Ninth Circuit practice. Mandatory AI disclosure on all filings for two years. The court stated the discipline was owed to the repeated failure of candor, not the initial error.
Court
U.S. District Court, District of Oregon
Judge
Magistrate Judge Mark D. Clarke
What Happened
Two attorneys representing a woman in a winery inheritance dispute submitted three briefs containing 15 references to nonexistent cases and 8 fabricated quotations. After opposing counsel identified the fabrications, attorney Stephen Brigandi removed the false information and refiled without disclosing the change to the court.
Outcome
$110,000 in fines and attorney fees. Largest AI hallucination sanction in US legal history. Case dismissed with prejudice. Both attorneys referred to the Oregon State Bar. The court found the conduct constituted an attempted cover-up.
Court
Nebraska Supreme Court
Judge
Chief Justice Michael Heavican
What Happened
Omaha attorney W. Gregory Lake used generative AI to draft a divorce appellate brief containing 57 defective citations out of 63, including 20 outright hallucinations and 3 entirely fabricated cases. He denied using AI when questioned during oral argument before the court, then reversed and admitted it after a suspension was recommended.
Outcome
Indefinite license suspension. First bar discipline action in US history to suspend an attorney's practice entirely over AI hallucination. A full disciplinary hearing was scheduled to determine the final term.
Court
U.S. Bankruptcy Court, S.D.N.Y.
Judge
Chief Judge Martin Glenn
What Happened
A partner at Sullivan and Cromwell, one of the most prestigious firms in the United States, filed an emergency motion in a Chapter 15 bankruptcy containing approximately 40 AI-generated errors: fabricated citations, misquoted authorities, wrong volume numbers, and quotations attributed to cases that did not contain them. Opposing counsel at Boies Schiller identified the errors.
Outcome
No sanctions imposed. The partner filed a three-page apology letter to the court. The firm acknowledged its own AI safeguards existed but were not followed. The outcome demonstrates that prompt candor and correction can avoid sanctions.
Court
U.S. District Court, District of Wyoming
Judge
Judge Kelly H. Rankin
What Happened
Three attorneys from Morgan and Morgan, the largest personal injury firm in the United States, filed a motion in limine citing nine cases. Eight did not exist. The citations were generated by the firm's internal AI platform. Two signing attorneys had not read the motion before it was filed.
Outcome
$5,000 total. Rudwin Ayala sanctioned $3,000 and removed from the case. T. Michael Morgan and Taly Goody sanctioned $1,000 each. The court held that signing a filing is non-delegable. The sanction later cost T. Michael Morgan a pro hac vice admission in an unrelated Massachusetts matter.
Court
U.S. Court of Appeals, Seventh Circuit
Judge
Seventh Circuit panel
What Happened
An attorney cited two nonexistent cases and included a false quotation in the standard-of-review section of an appellate brief. Counsel denied AI use at oral argument, then admitted the citations were copied from an unlocatable brief. First Seventh Circuit ruling to address AI hallucination and to clarify opposing counsel's duty to investigate suspect citations.
Outcome
Admonishment. No financial sanction. The court declined further penalty because the errors were unintentional, counsel was contrite, and the hallucinations supported an undisputed legal standard.
Court
U.S. District Court, N.D. California
Judge
Magistrate Judge Nathanael M. Cousins
What Happened
Plaintiff's counsel submitted a second motion for preliminary approval of a wage and hour collective action settlement containing AI-hallucinated citations. The fabrications in a settlement approval motion undermined the court's ability to evaluate the settlement on accurate legal grounds.
Outcome
$250 sanction to the clerk of court. Motion struck without leave to refile. Referral to the court's Standing Committee on Professional Conduct. Final settlement disposition delayed by months.
Court
U.S. District Court, S.D.N.Y.
Judge
Judge P. Kevin Castel
What Happened
Attorneys submitted a brief in a personal injury action citing six cases generated by ChatGPT that did not exist. When the court requested copies, the attorneys submitted additional fabricated case summaries, also AI-generated. The first widely publicized AI citation hallucination sanction in US courts.
Outcome
$5,000 sanction. The attorneys were ordered to notify each judge falsely identified as the author of a fabricated opinion. Established the baseline principle: an attorney who uses AI retains responsibility for verifying every citation.
What the Sanctions Establish
5Three principles recur across every ruling. First, the duty to verify is non-delegable. An attorney who signs a filing is responsible for its citations regardless of who drafted it, and regardless of whether AI produced them. Second, candor matters more than the initial error. The Ninth Circuit in LNU v. Blanche imposed suspension for the failure to disclose AI use, not for the use itself. Sullivan and Cromwell avoided sanction entirely through prompt disclosure. Third, commercial AI platforms are not a safe harbor. Sanctioned attorneys used internal firm AI, paid legal research tools, and consumer chatbots alike.
The Architectural Distinction
6A system that retrieves citations from a verified index cannot return a citation that does not exist. A system that generates text can. The difference between the two is not a matter of model quality or prompt engineering. It is a matter of architecture. The sanctions on this page document what happens when generated citations reach a court. The architecture that prevents it is retrieval-first constraint.