Sullivan and Cromwell Apologises to Bankruptcy Judge Over 40 AI-Hallucinated Citations
On April 8, 2026, Sullivan and Cromwell filed an emergency motion in the US Bankruptcy Court for the Southern District of New York in the case of Prince Global Holdings. The filing contained approximately 40 incorrect legal citations — wrong case names, wrong docket numbers, inaccurate article titles, and in some cases entire sentences that appear to have been AI-generated.
Opposing counsel at Boies Schiller Flexner spotted the errors and alerted the firm. On April 18, Andrew Dietderich, co-head of Sullivan and Cromwell’s global restructuring practice, sent a letter of apology to Chief Bankruptcy Judge Martin Glenn. He attached a red-lined version of the motion correcting the errors and acknowledged the firm had not followed its own AI review protocols.
“The Firm and I are keenly aware of our responsibility to ensure the accuracy of all submissions,” Dietderich wrote. “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 Case
Prince Global Holdings is a Cambodian conglomerate whose founder, Chen Zhi, faces charges of orchestrating forced labour operations and large-scale investment fraud. The bankruptcy is before Judge Glenn in SDNY (case 1:26-bk-10769). The emergency motion in question was part of the active wind-down proceedings.
Sullivan and Cromwell has confirmed it is conducting an internal investigation and “evaluating whether further enhancements to its internal training and review processes are warranted.” A hearing was scheduled for the following Wednesday with the corrected filing on the agenda.
The Broader Problem
This is not an isolated incident. A database maintained by legal technologist Damien Charlotin has recorded 1,334 AI hallucination incidents in court filings worldwide — more than 900 in the United States alone. Judges in some cases have imposed financial penalties.
What distinguishes the Sullivan and Cromwell incident is the institution’s profile. This is one of the most prestigious restructuring practices on Wall Street, not a solo practitioner filing from a general template. The firm had AI usage policies in place. The policies simply were not followed.
That distinction matters more than the error count. AI hallucination risk in legal work is no longer a question of whether firms have policies — most of the 1,334+ incidents involve firms that had policies. The question is whether review processes are enforced under deadline pressure, where AI tools are most likely to be used unsupervised and errors most likely to slip through.
The corrected motion has been accepted. The court has not indicated sanctions proceedings.