AI Hallucination Cases

Rodgers' declaration included a non-existent Wisconsin criminal case citation (reported by Bloomberg Law). Court found the error inadvertent, attributable to reliance on Bloomberg, and denied sanctions because conduct was neither reckless nor in bad faith.

Appellant Father's counsel submitted an opening brief containing misleading/inaccurate legal citations, mischaracterized case law, fabricated quotations attributed to Owen v. Blackhawk, and false quotations purportedly from the amended decree and hearing transcripts. The appellate court identified these errors, concluded they supported a potential ethics referral, and forwarded the decision to the State Bar for review.

Report on Recommendations can be found here.

The plaintiff's sworn declaration contained multiple fabricated quotations and manufactured deposition citations. Defendants identified the fabrications in a motion to strike; the court found the declaration filed in bad faith, struck it, and imposed sanctions under Fed. R. Civ. P. 56(h). The plaintiff admitted using generative AI to draft portions and failing to review the declaration; counsel failed to verify the citations despite attending and taking the depositions.

By interim order dated March 17, 2025 the court struck defendants' opposition and cross-motion after finding the filings, drafted with AI, contained hallucinated (fabricated) case citations. The court treated the motion as unopposed thereafter.

See story here.

The court eventually granted some costs in a subsequent decision (see here).

The court also identified five other cases in which the same firm or attorney confessed having misused AI. This despite the attorneys attending CLE training

MIO produced a prolix Defence containing multiple non-existent, miscited and misapplied authorities. The Court found AI-driven research produced hallucinatory results and that MIO failed adequately to verify authorities, amounting to reckless conduct warranting an indemnity costs order.

The court found that the majority of legal authorities in appellant counsel's reply brief were incorrect or did not support the propositions for which they were cited. The court struck the reply brief, imposed monetary sanctions of $5,070, and directed a copy of the opinion be forwarded to the State Bar.

Counsel had already been sanctioned in different case, and professed having gone through CLE on generative AI.

"Unfortunately, we're finding this problem arising more and more frequently […] When a lawyer cites imaginary legal authorities to our court as if they were law, we are compelled to refer that lawyer to the Bar because of the professional rules of conduct.

It doesn't take much moral imagination to understand why. As judges, we rely on attorneys to ethically represent their clients. We expect that representation to be zealous, honest, and competent. Indeed, lawyers owe the courts and their clients a duty to practice with competence and candor. By signing an appellate brief, a lawyer certifies that he or she has read the document and that to the best of the lawyer's knowledge, information, and belief there are "good grounds to support the document."

These ethical requirements are not excused simply because a computer program generated a faulty or misleading legal analysis. Nor is it an excuse that the attorney did not intend to mislead the court. "To state the obvious, it is a fundamental duty of attorneys to read the legal authorities they cite in appellate briefs or any other court filings to determine that the authorities stand for the propositions for which they are cited."

Obviously, that didn't happen when Ms. McLane filed this answer brief. Instead, counsel "fundamentally abdicated" her duty to the court and her client when she submitted this filing without verifying that the three cases cited in her brief said what she claimed they said. Accordingly, it is our duty to refer this matter to the Florida Bar to proceed as it deems appropriate."

See story here and here.