As posted to Linked In
Alvin Antony
Judges are using AI and they have a lot to say about it.
The AI Policy Consortium for Law and Courts, a joint initiative of the National Center for State Courts and the Thomson Reuters Institute, released a report titled “Judicial Use of Generative AI: Lessons Learned.” It is based on in-depth interviews with 13 state and federal judges across 10 US states.
Every single judge interviewed was already using GenAI in some form. But every single one also agreed on one thing: judges must always remain the final decision makers. Several judges said they do not even consult GenAI until after they have independently decided a case outcome. Many compared using GenAI to working with a judicial intern, useful for support, never for judgment.
The biggest benefit judges reported was efficiency. They are using GenAI to draft speeches, summarize lengthy filings, prepare for oral arguments, plan trainings, and even transcribe committee meetings. A few are exploring using it to improve access to justice for self-represented litigants and to simplify court communications for the general public.
But the risks kept judges up at night too. Hallucinations were the top concern, nearly every judge had personally experienced a GenAI tool producing a fictitious or misleading case citation. Privacy and cybersecurity were close behind, with judges noting that even their prompt history could inadvertently reveal their thinking on a pending case. Judges also worried about deskilling among young lawyers, negative public perception of AI in courtrooms, and a flood of low-quality filings from litigants misusing GenAI tools.
The judges interviewed also flagged that GenAI tends to express things with false certainty, can be sycophantic, and may carry bias from its training data, all serious concerns in a judicial setting.
The most used tool? ChatGPT, mentioned more than any other.
A copy of the report is enclosed with this post.
Read & Download Report
