AI Hallucinations in Filing by a Top Law Firm

David Lat (Original Jurisdiction), whose newsletter I very much recommend, writes about this:

Tonight brings huge news out of Sullivan & Cromwell. We can now add S&C, truly one of the world’s most prestigious and profitable firms, to the AI “Hall of Shame”: the list of law firms that have submitted court filings containing AI-generated “hallucinations.” …

[Ironically], as noted by Joe Patrice of Above the Law, Sullivan & Cromwell is the law firm that … advises OpenAI on the “safe and ethical deployment” of artificial intelligence (a representation S&C touts on its website). Physician, heal thyself?

[And] the AI errors were called to [Dietderich’s] attention by… Boies Schiller Flexner. Why is it ironic that BSF, one of the nation’s leading litigation firms, caught these mistakes? Because until now, Boies Schiller was the reigning champion of “Elite Law Firms Guilty of AI Fails.” (In addition to the AI snafu chronicled in that story, which took place last year, BSF subsequently failed to catch AI-generated errors in a filing by co-counsel in a different case.) …

When I covered BSF’s AI error last year, I gave props to then-partner John Kucera: he “fully and freely admitted the errors, and he didn’t throw any colleagues under the bus.” I’ll say the same here about Andy Dietderich. He’s a senior partner at S&C, having joined the firm almost 30 years ago; he leads its restructuring practice, which he fonded; and he’s a giant of the bankruptcy bar, Chambers Band 1. Nobody was paying him $2,500 an hour to do legal research; this screw-up was clearly the handiwork of an associate (or perhaps I should say former associate). But Dietderich signed the letter solo, and he didn’t point the finger at anyone else.

Here’s the substance of Dietderich’s letter:

I became aware Thursday evening that the JPLs’ Emergency Motion for Entry of Orders Granting (I) Ex Parte Relief and (II) Provisional Relief, Pursuant to Section 1519 of the Bankruptcy Code filed with the Court on April 9, 2026 (the “Motion”) includes inaccurate citations and other errors, each of which is identified and corrected in Schedule A to this letter.

The inaccuracies and errors in the Motion include artificial intelligence (“AI”) “hallucinations.” “Hallucinations” are instances in which artificial intelligence tools fabricate case citations, misquote authorities, or generate non-existent legal sources. We deeply regret that this has occurred. The Firm maintains comprehensive policies and training requirements governing the use of AI tools in legal work. These safeguards are designed to prevent exactly this situation. The Firm’s policies on the use of AI were not followed in connection with the preparation of the Motion. In addition, the Firm has general policies and training requirements for the proper review of legal citations. Regrettably, this review process did not identify the inaccurate citations generated by AI, nor did it identify other errors that appear to have resulted in whole or in part from manual error.

The Firm has undertaken immediate remedial measures, including but not limited to a full review of the circumstances leading to these errors, and a re-review of all filings in this matter, and we can confirm that the other filings do not include any AI-related errors. Our re-review did identify non-substantive and/or clerical errors in other filings in this matter (and one citation that confuses Section 1517(d) and Rule 6004(h)), which we do not believe to be the result of AI usage. These errors are also identified in Schedule A for the sake of completeness.

The Firm and I are keenly aware of our responsibility to ensure the accuracy of all submissions including under Local Bankruptcy Rule 9011-1(d), and I take responsibility for the failure to do so.

I want to assure the Court that the Firm’s policies governing AI use are both clear and rigorous. Access to AI tools is conditioned on completion of mandatory training. Before any Firm lawyer is granted access to generative AI tools, the lawyer must complete two required training modules, completion of which is tracked and verified.

The training repeatedly emphasizes the risk of AI “hallucinations,” including the fabrication of case citations, misinterpretation of authorities, and inaccurate quotations. It instructs lawyers to “trust nothing and verify everything” and makes clear that failure to independently verify AI-generated output constitutes a violation of Firm policy. The training also reviews the significant consequences of AI-related errors in various cases.

These requirements are reinforced in the Firm’s Office Manual for Lawyers, which provides that lawyers “must independently check all answers, case citations, and other information or work product received from an AI Program for both substantive and non-substantive accuracy.” The policy further states that no communication may be sent to a court, regulator, client, or other external party without the exercise of appropriate professional judgment and oversight.

Notwithstanding these safeguards, the Firm’s protocols were not followed here. The Firm is also evaluating whether further enhancements to its internal training and review processes are warranted.

We sincerely regret the errors in the Motion and the burden they have imposed on the Court and the parties, and I apologize on behalf of our entire team. I also called Boies Schiller Flexner LLP on Friday to thank them for bringing this matter to our attention and to apologize directly to them as well. A corrected version of the Motion, together with a redline reflecting the revisions, is being filed later today.