Writing in May for Above the Law, Ken Crutchfield observed that “over the past year, we’ve seen a steady stream of headlines about AI hallucinations where fake cases, fabricated quotes, and misstated facts made it through a review process and into court filings.” The predicament isn’t just that GenAI tools invent non-existent cases that some short-cut-taking attorneys cite without verifying; it’s that the tools also produce difficult-to-discover false claims and quotations that masquerade as truth in pleadings and motions. Indeed, the US Court of Appeals for the Fifth Circuit explained in February that the “problem now often manifests as false quotes or statements of law attributed to real cases, rather than the more easily recognizable fake cases.”
I first examined these troubles more than three years ago, suggesting that in terms of
bottom-line human responsibility, it helps to recall that good senior partners in charge of a case—the lead attorneys—should always review the work of junior associates they assign to it. That same sense of personal, buck-stops-here accountability must apply whenever any attorney farms out work to a generative AI system.
I also addressed the problem in January 2024 and again in May 2025 because some attorneys didn’t learn from the earlier mistakes of their colleagues. Unfortunately, this post now makes good on a prediction made by AEI’s technology policy scholars this January: That in 2026, “[a]n attorney somewhere in the US will cite output from a generative artificial intelligence tool that is erroneous, and we’ll once again need to remind lawyers of this danger.”
Several things, however, are different than they were three years ago: (1) a body of appellate court decisions addressing the issue is emerging; (2) state supreme courts and bar associations are issuing new guidelines, best practices, and amended rules; and (3) some courts’ deep disappointment with attorneys who cite hallucinated GenAI output is palpable. Helpfully, Ropes & Gray provides a collection of orders, rules, and decisions about AI usage.
Sometimes white-shoe law firms slip up. The New York Times reported in April that Sullivan & Cromwell “apologized to a federal judge for submitting a court filing replete with errors created by artificial intelligence, including ‘hallucinations’ that fabricated case citations.” Andrew G. Dietderich, a partner at the firm, wrote a letter to Chief Judge Martin Glenn of the US Bankruptcy Court for the Southern District of New York apologizing and expressing deep “regret that this has occurred.”
An attorney at a well-known, national personal injury firm recently faced repercussions. In May, Justice Kenneth Salinger of the Suffolk County (Mass.) Superior Court denied Florida attorney T. Michael Morgan’s motion to appear pro hac vice in a case before Salinger. The justice noted that Morgan (of Morgan & Morgan) was sanctioned in 2025 by a federal judge in Wyoming after he “permitted the filing of motions citing eight non-existent cases that had been hallucinated by Morgan & Morgan’s in-house Artificial Intelligence (‘AI’) platform.” Salinger was “troubled by Morgan’s demonstrated failure in the Wyoming case to live up to the ethical standards required of trial lawyers,” adding that “Morgan has made no showing that he has learned from his mistakes in that case.” In short, failings with GenAI from the recent past in one jurisdiction can haunt an attorney later and elsewhere.
Reviewing appellate court rulings is useful. The US Court of Appeals for the Sixth Circuit’s April 2026 ruling in United States v. Farris stresses that “the legal profession must be clear eyed about technology’s potential pitfalls. That mandate is especially critical in today’s rapidly evolving artificial-intelligence landscape.” The opinion notes that “[a]ttorneys have an ethical obligation to verify the citations and propositions they submit to courts; that obligation reflects duties of competence and candor that apply no matter the tools attorneys use.” The court usefully provides a non-exhaustive list of six “relevant steps” attorneys should take to fulfill their ethical duties when using AI tools.
The Fifth Circuit’s February 2026 opinion in Fletcher v. Experian Information Solutions is instructive. It suggests there’s no excuse for attorneys to plead ignorance about hallucination problems when using GenAI tools, given “numerous news stories, CLE [continuing legal education] presentations, scholarly articles, and judicial entreaties” about these matters. It also notes that “submitting a brief riddled with fabricated quotations and assertions” is a sanctionable abuse of the judicial process. Furthermore, the court called out the attorney in question for not accepting responsibility for her actions: “when confronted with a serious ethical misstep, [she] misled, evaded, and violated her duties as an officer of this court.”
The solution here for attorneys involves a basic yet vital task—independently verifying all cases, facts, and quotations generated by AI tools. As Crutchfield explains, this entails “a complete review of everything, including foundational facts,” at the end of the drafting and editing process.