FOLLOW US

Sullivan & Cromwell — one of America's most powerful and prestigious law firms — has been forced to apologize to a federal judge after artificial intelligence hallucinations contaminated a major court filing in the high-profile Prince Group case, misquoting federal law and fabricating case citations.

BY SAMUEL A. LÓPEZ ◆ USA HERALD ◆ INVESTIGATIVE DESK ◆ LEGAL & NATIONAL SECURITY

Here's something that should make every client of a major law firm stop and think twice: Sullivan & Cromwell — a firm with more than 900 lawyers and one of the most storied reputations in American corporate law — just admitted to a federal judge that artificial intelligence hallucinated its way into one of its court filings. And it wasn't caught by S&C's own team. It was caught by opposing counsel.

The firm's co-head of global restructuring, Andrew Dietderich, sent a letter on Saturday to U.S. Bankruptcy Judge Martin Glenn in New York, issuing a direct apology for a cascade of errors found in a filing dated April 9.

The mistakes — which included misquoting the US bankruptcy code and citing cases that were inaccurately summarized or incorrectly attributed — were the product of AI-generated content that slipped through what the firm now admits was a failed review process.

"We deeply regret that this has occurred. I apologize on behalf of our entire team."

— ANDREW DIETDERICH, CO-HEAD OF GLOBAL RESTRUCTURING, SULLIVAN & CROMWELL

Let that sink in. A firm that charges some of the highest billing rates in the world — a firm that corporations, governments, and billionaires trust with their most critical legal battles — let an AI system fabricate legal citations in a federal court filing. And no one on the S&C team caught it.

WHAT HAPPENED

The errors were uncovered by the law firm Boies Schiller Flexner (BSF), which was also working on the case. BSF flagged multiple instances in which S&C had filed inaccurately summarized conclusions drawn from other cases — the kind of subtle, authoritative-sounding distortions that AI language models are notorious for producing.

In his letter, Dietderich acknowledged that S&C maintains "comprehensive policies and training requirements governing the use of AI tools in legal work" — safeguards specifically designed to prevent exactly this kind of catastrophic error.

The problem? Those policies were not followed. And a secondary review process, which should have served as a backstop, also failed to catch the inaccurate AI-generated citations.

⚠ CRITICAL DISCLOSURE

The letter did not identify which AI program was used, which lawyers prepared the filing, or whether any internal disciplinary action has been taken against those responsible. These are questions the public — and every client of Sullivan & Cromwell — deserves answered.

S&C later filed a corrected version with the court. Lawyers are not prohibited from using AI tools, but under professional ethics rules, they are absolutely bound to ensure the accuracy and integrity of every document submitted to a court. There are no exceptions. There is no "the AI made me do it" defense.

THE CASE: PRINCE GROUP & CHEN ZHI