Two US federal judges admit AI use caused errors in court rulings
Two US federal judges admitted that their staff improperly used AI tools including ChatGPT and Perplexity to draft court orders containing serious factual errors. The judges said the orders bypassed normal review processes, were corrected after the errors were identified, and they subsequently implemented stronger AI governance policies.
What happened
Two US federal judges admitted that their staff improperly used AI tools to help draft recent court orders that contained serious errors. Judge Henry Wingate in Mississippi said a law clerk used Perplexity as a foundational drafting assistant for a civil rights lawsuit ruling, with posting the draft decision described as a lapse in human oversight. Judge Julien Xavier Neals in New Jersey said a law school intern used OpenAI's ChatGPT for research without authorization or disclosure in a securities lawsuit decision that was released in error and subsequently withdrawn.
Both judges stated the rulings did not go through their chambers' typical review processes before issuance. The judges made these admissions in letters responding to an inquiry by US Senate Judiciary Committee Chairman Chuck Grassley after lawyers in the cases reported the decisions contained factual inaccuracies and other serious errors.
Reported impact
Lawyers in the cases identified factual inaccuracies and other serious errors in the court rulings. Judge Neals' securities lawsuit decision was released in error and withdrawn as soon as it was brought to the attention of chambers.
Index analysis
Both judges have since adopted corrective measures. Judge Neals' chambers created a written AI policy and enhanced its review process, while Judge Wingate's chamber implemented new oversight procedures.
Governance controls
Grassley stated that each federal judge and the judiciary as an institution has an obligation to ensure generative AI use does not violate litigants' rights or prevent fair treatment under law, and he urged the judiciary to adopt stronger AI guidelines.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.