Top law firm Sullivan & Cromwell told a US federal bankruptcy court that a major filing it made in a high-profile case contained multiple AI hallucinations
Firm whose partners bill more than $2,000 per hour apologises to judge for software-driven errors in bankruptcy case
Context & Ripple Effects
Related coverage has tracked AI moving into work traditionally handled by junior lawyers, alongside repeated court warnings over fabricated authorities and other AI-generated legal material.
The incident also lands at Sullivan & Cromwell, a firm deeply involved in the FTX bankruptcy aftermath, where the reliability of major filings carries particular weight.
First-order effects
- Sullivan & Cromwell must account to the bankruptcy court for errors in a major filing, putting its review process and the filing’s reliability under immediate scrutiny.
- The court and parties in the case may need to spend additional time validating the affected legal material rather than relying on the submission as filed.
Second-order effects
- Large firms deploying AI in research and drafting will face pressure to add or tighten lawyer-led citation and source verification, especially for court-bound work.
- Clients paying premium rates may more directly test whether AI-assisted workflows lower cost or instead introduce review, delay, and reputational risk.
Third-order effects
- If such incidents persist, AI use in legal practice is likely to be governed less by tool adoption alone and more by auditable supervision standards, with professional responsibility remaining with the lawyers signing filings.
- The efficiency case for automating entry-level legal work will depend on whether firms can prove that human verification scales reliably enough to prevent costly court failures.
The trend: Generative AI is shifting legal work from an adoption question to a governance and accountability question, as courts test whether firms’ controls can match the technology’s use in high-stakes filings.