AI in the Federal Courts: Inside the Judges’ Survey

In this episode from the City Bar’s Artificial Intelligence and Digital Technologies Task Force, host David Zaslowsky is joined by the Honorable Xavier Rodriguez, a United States District Judge for the Western District of Texas and a co-founder of the Judicial AI Consortium, to dive deep into a random-sample survey of 502 federal judges on their use of AI, conducted by Judge Rodriguez and researchers at Northwestern University in collaboration with the City Bar. According to the survey, more than 60% of federal judges have used an AI tool in their work, but only 22% use one weekly or daily. Judges favor legally-curated tools over general-purpose platforms, and legal research dominates the use cases. Most judges have never been offered AI training. One in three permits AI use in chambers; one in four has no policy at all. David and Judge Rodriguez discuss why adoption is broad but infrequent, why chambers without an explicit policy will get shadow use regardless, and the growing number of court orders found to contain AI-generated errors. The conversation closes on an in-house AI server experiment for Social Security appeals and the principle of high value, low risk. Read the survey, Artificial Intelligence in Federal Courts: A Random Sample Survey of Judges, published by the Sedona Conference: https://www.lawnext.com/wp-content/uploads/2026/03/Artificial_Intelligence_in_Federal_Courts_preprint.pdf 01:51 From E-Discovery to the Judicial AI Consortium 04:26 How the Survey Was Built 07:00 Six Insights from the Data 11:09 What Surprised a Judge 12:45 Shadow Use and Missing Policies 13:42 Potential Biases in the Study 17:40 What AI Training Should Cover 22:53 Inside the Judicial AI Consortium 26:35 When Court Orders Contain AI Errors 33:28 Hallucinating Before AI 36:49 A Better Product for Litigants 39:26 Optimism, Concern, and Adoption 43:34 An In-House AI Server Experiment 47:03 High Value, Low Risk