- First widely-publicized AI hallucination sanctions order, issued June 22 2023.
- $5,000 monetary sanction against two attorneys and their firm under FRCP 11(b), jointly and severally.
- Court was explicit that AI use is not itself improper. The sanction was for the failure to verify.
- Reputational consequences ran far beyond the dollar penalty. The names now anchor every subsequent AI-sanctions discussion.
Mata v. Avianca, Inc..
An attorney who files a brief containing AI-generated fabricated case citations violates FRCP 11(b) regardless of subjective good faith, because the rule's "reasonable inquiry" element requires the attorney to verify each cited authority exists and supports the proposition for which it is offered.
What happened. #
Counsel for the plaintiff in a routine personal-injury case against Avianca filed an opposition brief that cited and quoted from federal cases that did not exist. The fabricated authorities had been generated by ChatGPT, which counsel used to research the brief and did not verify against any primary database. When opposing counsel could not locate the cited cases, the court issued an order to show cause. The attorneys admitted the use of AI, admitted that they had not verified the citations, and submitted that they had asked ChatGPT whether the cases were real and accepted its affirmation.
The holding. #
An attorney who files a brief containing AI-generated fabricated case citations violates FRCP 11(b) regardless of subjective good faith, because the rule's "reasonable inquiry" element requires the attorney to verify each cited authority exists and supports the proposition for which it is offered.
The court's reasoning. #
Judge Castel's sanctions order grounded the sanction in FRCP 11(b), which requires that, by signing a paper presented to the court, an attorney certifies "that to the best of the person's knowledge, information, and belief, formed after an inquiry reasonable under the circumstances," the legal contentions are warranted by existing law. The "reasonable inquiry" element, the court held, has always required reading the cases one cites. The introduction of an AI tool between the attorney and the authority does not transfer the duty. The attorney still signs the brief. The signature still carries the certification.
Operational implications. #
For firms operationalising AI-assisted research, Mata stands for two concrete propositions. First, every cited authority must be independently verified in a primary database before filing. Second, the firm's AI policy must memorialise that verification as a documented step, not an assumed one. The procedural defence to a future sanctions inquiry is the firm's written workflow that shows each citation was checked, by whom, and when. Engagement with the firm's state-bar AI ethics opinion (and where applicable, ABA Formal Opinion 512) is foundational to that workflow.
Primary sources. #
- Mata v. Avianca opinion, CourtListener
- FRCP 11, Legal Information Institute
- ABA Formal Opinion 512, ABA
This is a case brief, not legal advice. Dan Hughes is not an attorney; IXSOR does not provide legal services.