AI Hallucination Cases

Plaintiff's counsel filed an opposition brief containing several quotations and citations the Court could not verify, including at least one the Court described as a "fabricated quotation." Defendants first identified the potentially inaccurate citations in their reply. The Court ordered disclosure about AI use, production of drafts, metadata and timesheets, and set a hearing to determine responsibility and whether sanctions or bar referral are warranted.

Show Cause Order here. Law firm explained what happened here.

In the ultimate order, the court noted that "In terms of competence, the threat to attorneys using generative artificial intelligence platforms powered by large language models is two-fold. First, danger exists that the attorney does not understand how the technology functions, believing that the output is real instead of “realistic-looking."."

In finding that the law firm acted with integrity, the Court noted with approval that it had repaid the other side's fees, to the tune of 55,721.2 USD.

Order to Show Cause is here.

Show Cause Order is here.

"The danger of Attorney Stich’s AI use is especially felt here because his opponent’s pro se status meant that there was not an adversary capable of calling the attention of the court to the phony citations. Furthermore, this court expended time and resources in investigating the hallucinated citations, resources that could have been better spent adjudicating the merits of this underlying litigation and that of other cases pending before this court.

The oversight in submitting fake citations is more than just sloppy lawyering: it imperils the integrity of our judicial process.However, the court also acknowledges that our society sits on the precipice of rapid technological development and that the continued development of AI will fundamentally alter life as we know it. Just as the advent of the Internet in the late 20th century transformed the legal profession, and particularly legal research, so too will artificial intelligence. Indeed, the two biggest legal research databases, Westlaw and LexisNexis, have developed and continue to expand their own proprietary AI tools to assist legal practitioners in finding case law.2 This Order should not be construed as a Luddite attack on technology and the efficiency it brings to the legal profession. Rather, this Order is an acknowledgement that AI remains a nascent technology with questionable reliability at this juncture. Given the ethical obligations lawyers must honor, it is imperative that lawyers use AI with diligence and care. This technology is too unsophisticated and must necessarily yield to a lawyer’s obligation of candor to the court."

From the Order to Show Cause (available here): "One week after filing its second motion for summary judgment, Liberty Mutual submitted a notice of errata identifying the erroneous Goodman and Chaudri citations and demonstrating legitimate citations to those cases. [ECF No. 50].3 While the Court acknowledges Liberty Mutual’s prompt notice disclosing the two most serious errors in its filing, the additional misquotations and mischaracterizations discussed above will not be disregarded. Liberty Mutual indicates that the errors were typographical and/or caused by vision impairment, but that explanation is simply not credible. The errors in Liberty Mutual’s filing are not ones in which a few letters or numbers were passed over or shuffled. Rather, the filing includes entire names, dates, court designations, and Westlaw citations that are completely off base, and various other inaccuracies cannot be explained by typographical or vision issues. Therefore, the Court will reserve its ruling on the motion for sanctions and will set a hearing requiring Liberty Mutual to show cause why it should not be sanctioned."

Later on, the court accepted Counsel's technical audit that suggested the errors stemmed from a human, non-AI source.

Defendants' reply identified fictitious cases, incorrect citations, and non-existent quotations in Plaintiffs' response brief. Plaintiffs filed a Notice of Corrected Citations; the Court noted formatting that suggested use of generative AI but declined to sanction the pro se plaintiffs, advising compliance with Rule 11 in future filings.

See beginning of judgment: "Editor's Note: This decision contains discussion of citation references that are incorrect or do not actually exist. These invalid citations appeared in the original court opinion and have been preserved as written since they are part of the official record. Any links to these invalid citations have been removed."

Fine was 3,000 USD, to which the court added 25,000 USD in costs in a subsequent order.

When warned about the issue, cousel apologised in an open letter to the Court.

In his response, Counsel acknowledged that he first researched relevant cases, and then "entered the cases into ChatGPT and requested that it highlight favorable arguments contained in the list of cases."

Plaintiff's opposition brief cited numerous nonexistent cases. Defendant identified the fabricated citations in its reply; plaintiff admitted reliance on an AI-based drafting tool and apologized. The court confirmed several cited cases do not exist but declined to pursue Rule 11 sanctions, finding apology, mitigation, and proportionality concerns.

The Court identified multiple instances where counsel cited nonexistent cases and misquoted Cal. Civ. Code § 988(c). The Court admonished counsel, noted corrections in later briefing, and declined to credit the arguments based on the erroneous citations. No sanctions were imposed.

Morehouse replied that Warfield's filings relied on incorrect and nonexistent case citations; the Court noted that allegation but did not impose sanctions and dismissed the case for lack of personal jurisdiction.