Back to home
Regulation
Sullivan & Cromwell apologises to federal judge over AI hallucinations in court filing
The elite law firm admitted fabricated content generated by AI tools appeared in a submission to a bankruptcy court.
Share
Briefing
- •Sullivan & Cromwell, one of Wall Street's most prominent law firms, apologised to a federal bankruptcy judge after AI-generated hallucinations produced errors in a court filing.
- •The firm acknowledged that AI tools introduced fabricated or inaccurate content into the submission, requiring a formal correction to the court.
- •The incident follows a pattern of law firms facing judicial sanctions for AI-generated citations and claims that cannot be verified or do not exist.
- •The episode raises direct questions about the supervision protocols governing AI use at firms that have publicly promoted legal tech adoption.
Why it matters
- →Sullivan & Cromwell's incident will accelerate court-mandated AI disclosure requirements across federal and bankruptcy courts. Firms that have publicly committed to legal tech adoption now face asymmetric reputational risk: any subsequent filing error attributed to AI triggers sanctions exposure and client credibility damage simultaneously. This raises the compliance cost of AI deployment for Am Law 100 firms, benefiting vendors that offer human-in-the-loop verification workflows over raw generative output.
- →Legal AI vendors targeting BigLaw, including Harvey, Casetext, and Thomson Reuters Practical Law, face a sales cycle slowdown as general counsel and managing partners pause or tighten deployment approvals following a high-profile hallucination incident at a firm of Sullivan & Cromwell's standing. The reputational stakes at Wall Street-facing firms are higher than at general practice firms because filing errors touch active deals, regulatory proceedings, and bankruptcy estates with quantifiable financial consequences.
Historical Context
- June 2023
Attorneys Mata v. Avianca submitted ChatGPT-generated citations that did not exist; Judge Castel sanctioned the lawyers involved. That ruling established that AI-generated errors are an attorney responsibility, not a technology defense, setting the precedent Sullivan & Cromwell now faces.
- 2020-2022
E-discovery vendors that overpromised automation accuracy in contract review faced widespread client clawbacks and reputational damage when error rates surfaced in litigation. The pattern of enterprise legal tech adoption followed by high-profile failure causing market-wide caution is a direct structural parallel.
