Law firm files AI-tainted motion
- A top Wall Street firm submitted a federal court filing that contained AI-generated errors and later apologised to the judge. - The filing included roughly 40 incorrect citations and other mistakes traced to AI-assisted drafting. - The episode shows generative AI can give authoritative-looking but false outputs, forcing stricter verification in professional workflows. (theguardian.com)
Sullivan & Cromwell told a New York bankruptcy judge that an April 9 emergency motion it filed contained artificial-intelligence errors, including fake and mangled legal citations. (reuters.com) The firm’s apology went to Chief Judge Martin Glenn of the U.S. Bankruptcy Court for the Southern District of New York on April 18, 2026. Andrew Dietderich, co-head of Sullivan & Cromwell’s restructuring group, said the filing had “inaccurate citations and other errors” generated with AI assistance. (reuters.com) The motion was filed in the Chapter 15 case of Prince Global Holdings Limited, a cross-border insolvency proceeding in Manhattan bankruptcy court. Public docket records show the case was filed on April 8 and the emergency motion notice was filed on April 9 under case number 26-10769 before Judge Glenn. (pacermonitor.com, pacermonitor.com) Reuters reported that the filing contained about 40 bad citations and other mistakes. Bloomberg Law reported the firm told the judge it had not properly reviewed the citations and was considering stronger internal training and review procedures. (reuters.com, bloomberglaw.com) In plain terms, an AI “hallucination” is when a chatbot produces text that looks polished and specific but is wrong or entirely invented. In legal work, that can mean a case that does not exist, a real case tied to the wrong quote, or a citation that points to the wrong page. (technologyreview.com, legaltechnology.com) This was not a filing from a solo practice or a small local shop. Sullivan & Cromwell is one of Wall Street’s best-known firms, and the mistake surfaced only after opposing counsel at Boies Schiller Flexner flagged the citation problems. (nytimes.com, legaltechnology.com) Courts have been confronting similar problems for three years. In 2023, lawyers in federal court in New York were sanctioned after filing a brief with nonexistent cases generated by ChatGPT, a case that became an early warning about unchecked AI in legal research. (nytimes.com, reuters.com) The judge in the Prince Global matter had not sanctioned Sullivan & Cromwell as of the latest reports on April 23. The firm’s apology said it “deeply regret[s]” the episode, and the case now stands as another test of whether law firms can use generative AI without letting invented authority reach a federal docket. (reuters.com, bloomberglaw.com)
