Top Wall Street Law Firm Apologizes to Judge for AI Hallucination in Court Filing

The law firm told a federal judge that a motion contained inaccurate citations and other mistakes linked in part to AI hallucinations.

Snow falls outside the New York Stock Exchange on Wall Street in New York City on Feb. 23, 2026. Seth Wenig/AP Photo

A prestigious Wall Street law firm has apologized to a federal judge for a filing that included erroneous content generated by artificial intelligence (AI).

Sullivan & Cromwell, the Manhattan-based white-shoe firm, sent a letter on April 18 to Chief Judge Martin Glenn of the U.S. Bankruptcy Court for the Southern District of New York, apologizing for submitting a motion that contained “inaccurate citations and other errors” caused in part by AI hallucinations, the phenomenon of AI chatbots generating plausible-sounding but false information.

“We deeply regret that this has occurred,” S&C partner Andrew Dietderich wrote in the letter.

Attached to the letter was a three-page list of roughly 40 corrections. The errors included incorrect citations, wrong volume numbers, and what appeared to be fabricated quotations attributed to real cases. Some of the mistakes, however, were clerical and not related to AI.

The flawed document arose in proceedings involving Prince Global Holdings, a Cambodian conglomerate whose chairman, Chen Zhi, has been charged by federal prosecutors with running an international fraud and money-laundering operation. Chen, who was not in the United States when the indictment was unsealed, was extradited from Cambodia to China in January.

According to Dietderich’s letter, he was alerted about the filing errors by lawyers at Boies Schiller Flexner, the law firm representing the Prince Group. After learning of the errors, S&C reviewed its other filings and concluded that AI hallucinations were confined to that single motion, Dietderich wrote.

Dietderich said that his firm requires lawyers to complete training before gaining access to AI tools and maintains “clear and rigorous” policies governing their use, including a basic instruction to “trust nothing and verify everything.”

“Notwithstanding these safeguards, the Firm’s protocols were not followed here,” he wrote, adding that the firm is considering improvements to its training and review procedures.

The apology adds to a growing list of AI-related blunders in the legal industry. Judges across the country have sanctioned attorneys who used generative AI for legal research and drafting but failed to verify the accuracy of the output.

In 2023, a federal judge in Manhattan imposed a $5,000 sanction after two lawyers and their firm submitted a brief containing six cases made up by ChatGPT.

More recently, in February, the U.S. Court of Appeals for the Fifth Circuit in New Orleans sanctioned an attorney $2,500 after finding that her brief contained 16 AI-fabricated quotations and five other “serious misrepresentations” of law or fact.

Damien Charlotin, a French lawyer and data scientist, maintains a public database tracking court decisions involving AI-hallucinated content. As of April 2026, the database has recorded more than 1,300 such cases across the world, including more than 900 in the United States.

“It is a problem that is getting worse—not better,” the Fifth Circuit said in the opinion, pointing to Charlotin’s findings. The court added that, if ignorance of the risks of using generative AI without vetting the results was ever an excuse, it is no longer valid.

“To ethically use generative AI in the practice of law, a lawyer must ensure that the legal propositions and authority generated are trustworthy,” the Fifth Circuit said.

A previous version of this article misspelled the name of Sullivan & Cromwell partner Andrew G. Dietderich. The Epoch Times regrets the error.