Federal Judges Admit AI-Fueled Errors, Order Reforms

Two U.S. district judges — Julien Xavier Neals (New Jersey) and Henry T. Wingate (Mississippi) — acknowledged that draft rulings their chambers signed contained factual errors and fabricated citations after staff used generative AI (Neals said a law‑school intern used ChatGPT; Wingate said a law clerk used Perplexity), and the flawed drafts were withdrawn. The admissions followed an inquiry by Senate Judiciary Committee Chairman Chuck Grassley, who commended their candor and urged stronger judiciary AI guidelines. Both judges said the problematic orders were drafts that escaped proper review, that no privileged or sealed material was disclosed, and that the staffers' use of AI violated chamber rules or occurred where no formal rules existed. Each judge has implemented corrective measures — including mandatory independent second reviews, printing and attaching cited cases, and clarifying or enforcing bans on unauthorized generative‑AI use — to prevent recurrence. The episodes highlight the broader problem of AI “hallucinations” in legal work and raise urgent questions about how courts will oversee and regulate generative AI.