Sullivan & Cromwell, one of the most powerful law firms on Wall Street, has apologized to a federal bankruptcy judge after a major court filing was found to contain fabricated citations, misquoted authorities, and other errors generated by artificial intelligence.

This is not a solo practitioner in a strip mall law office. This is a 140-year-old firm with more than 1,000 attorneys and one of the top reputations for corporate work in the United States. And its AI tools invented legal citations and submitted them to a federal court.

What Happened

Andrew Dietderich, co-head of Sullivan & Cromwell's Global Finance & Restructuring group, wrote a letter to Manhattan-based Chief Judge Martin Glenn on April 18, explaining that a filing made on April 9 contained inaccurate citations and AI hallucinations.

The errors were not caught internally. They were discovered by opposing counsel at Boies Schiller Flexner, which was also working on the case. A chart attached to Dietderich's letter showed the motion contained incorrect case names, wrong case numbers, and apparently fabricated quotes from legal decisions.

"We deeply regret that this has occurred," Dietderich wrote. "I take responsibility for the failure to do so."

The Double Failure

What makes this case particularly damaging is the double failure. Sullivan & Cromwell told the court it maintains "comprehensive policies and training requirements governing the use of AI tools in legal work" specifically designed to catch these errors. Those policies were not followed. A secondary review process also "did not identify the inaccurate citations generated by AI."

In other words: the firm had the safeguards. The safeguards failed. And the backup to the safeguards also failed. The errors only surfaced because opposing counsel checked the citations manually.

The letter did not name which AI tool was used, which lawyers prepared the document, or whether anyone faced internal consequences.

The Case Behind the Filing

The filing was part of Sullivan & Cromwell's representation of liquidators appointed by legal authorities in the British Virgin Islands in actions against Prince Group, owned by the Chinese-born businessman Chen Zhi. US prosecutors charged Chen with wire fraud and money laundering, alleging he directed forced-labor scam compounds across Cambodia that "stole billions of dollars from victims in the United States and around the world." Chen was arrested in Cambodia earlier this year and extradited to China.

A high-stakes international fraud case is exactly the kind of proceeding where fabricated citations can cause real damage to the court's ability to make sound decisions.

The Pattern That Will Not Stop

Sullivan & Cromwell is far from the first firm to get caught. Legal researcher Damien Charlotin maintains a public database tracking AI hallucinations in court filings, and the incidents have been climbing since 2023. Lawyers are not prohibited from using AI. They are ethically bound to verify the accuracy of everything they submit.

But this is the first time a firm of this caliber has been caught. When a solo practitioner submits fake citations, it is embarrassing. When Sullivan & Cromwell does it, it forces a question the entire legal profession has been trying to avoid: if the best-resourced firm in America, with dedicated AI policies and review processes, still cannot catch hallucinations before they reach a judge, who can?

The answer, for now, appears to be: opposing counsel. Which means the reliability of AI-assisted legal work depends entirely on whether someone on the other side is still reading the footnotes.

Sources: The Guardian, Business Insider, The New York Times, Sullivan & Cromwell court letter (April 18, 2026).