What happened with Sullivan & Cromwell’s AI-related legal error?

Sullivan & Cromwell submitted a court filing containing about 40 incorrect citations caused by AI hallucinations and oversight failures. The errors appeared in a US Bankruptcy Court filing, prompting the firm to apologise to the judge and acknowledge responsibility for the mistake.

Why did Sullivan & Cromwell apologise to the court?

The firm apologised because inaccurate AI-generated citations were included in an official legal filing, which breaches professional standards. In legal work, accuracy is critical, and submitting false or fabricated references can undermine both the case and the firm’s credibility.

What are AI hallucinations in legal filings?

AI hallucinations refer to instances where AI tools generate false or fabricated information, such as non-existent legal citations. In this case, the AI produced incorrect references that were not properly verified before submission.

How did internal controls fail at Sullivan & Cromwell?

Sullivan & Cromwell said it already had policies requiring verification of AI-generated content, including citation checks. However, those procedures were not followed, allowing the errors to pass through the review process and into the final filing.

How common are AI errors in court filings?

AI-related legal errors are becoming more frequent, with over 1,300 documented incidents globally, including more than 900 in the US. Most cases involve fabricated citations, showing a recurring risk when AI outputs are not carefully checked.

Why does this matter for investors and markets?

This incident highlights operational risk tied to AI adoption in professional services like law and finance. Firms using AI without strong controls could face legal, reputational and financial consequences, which may affect client trust and long-term business value.

What actions is Sullivan & Cromwell taking after the incident?

The firm has launched an internal investigation and implemented immediate remedial measures. It is also reviewing whether to strengthen its training and verification processes to prevent similar AI-related errors in future filings.

How does this relate to the broader AI adoption trend?

Many firms are rapidly integrating AI tools to improve efficiency, but this case shows the risks of relying on AI without proper human oversight. It reinforces the need for strict validation processes, especially in high-stakes industries like law.

Sullivan & Cromwell apologises over AI legal errors

Written by Jon CuthbertPublished Apr. 22 2026Last Updated Apr. 22 2026

Sullivan & Cromwell has apologised to a federal judge after submitting a legal filing containing about 40 incorrect citations and errors linked to AI hallucinations.

The firm said the mistakes appeared in an emergency motion filed in the US Bankruptcy Court for the Southern District of New York, raising concerns about oversight in high-stakes legal work.

“We deeply regret that this has occurred,”

Said Andrew Dietderich, co-head of the firm’s global restructuring team, in a letter to Chief Judge Martin Glenn.

Dietderich added that while the firm has policies requiring verification of AI-generated content, those procedures were not followed in this instance, allowing errors to pass through review.

The case adds to a growing number of AI-related legal missteps, with legal technologist Damien Charlotin tracking more than 1,300 incidents globally, most involving fabricated citations.

Sullivan & Cromwell, which previously represented FTX in its bankruptcy, said it has launched an internal investigation and implemented immediate remedial measures.

The firm is now reviewing whether to strengthen training and verification processes after a rival law firm identified the errors and alerted the court.