Judges identified as early adopters of GenAI are finding the technology's biggest benefit to be increased efficiency with administrative work, like planning continuing legal education events and drafting speeches, the National Center for State Courts said in the report released Friday.
Some judges are using GenAI to help summarize lengthy filings or draft written work, as well, according to "Judicial Use of Generative AI: Lessons Learned."
But nearly all of the 13 state and federal judges interviewed across 10 states also expressed concerns about the possibility that GenAI could hallucinate fictitious cases, risk privacy and cybersecurity, or spark an avalanche of filings from self-represented litigants, the report said.
"A popular analogy used by multiple judges was that they treat GenAI like a judicial intern, and only 'delegate' questions where they already know the general answer and are just looking for additional support," the report's authors said.
The early-adopting judges interviewed in October and November by the AI Policy Consortium for Law and Courts, a partnership between the NCSC and the Thomson Reuters Institute, reported using GenAI mainly on "lower-risk" tasks. These include generating ideas for speeches, planning CLE events, updating staff trainings and transcribing meeting notes.
Some of the jurists have experimented as well with asking the technology to take on work more traditionally associated with "judging," such as summarizing documents, brainstorming possible questions to ask litigants during oral arguments and drafting opinions, though only after independently deciding the case, the report said.
A few judges are also exploring how the technology can increase access to justice, for instance by providing self-represented litigants with guidance, making court rules more understandable, updating forms, and helping with sealing and expungement policies, according to the approximately 60-minute-long interviews with the anonymous jurists.
"I was surprised by the ways some court systems are experimenting with using GenAI to help the public better understand the work they are doing, such as by creating first drafts of summaries of opinions to share with the public," said Amy Cyphert, one of the report's co-authors and a professor at West Virginia University College of Law.
But "the judges we spoke with are very clear-eyed about both the risks and opportunities associated with GenAI use and they are committed to preserving the human elements of judging," Cyphert added.
Hallucinations seem to be the biggest concern for most judges, many of whom reported being directed by an AI tool to a fictitious case or a mistaken description of an actual case.
"Judges routinely stressed how important human oversight is, advising that judges must verify the accuracy of any GenAI output," according to the report. "Several judges noted that this need for oversight is in tension with the benefit of GenAI as a time saver, and that they do not use GenAI for legal research because of this: they ultimately felt it did not save them time in the end."
Judges are also keenly aware of the privacy and cybersecurity risks that come with GenAI, particularly in sensitive cases, such as those involving juveniles. Some of the interviewed jurists said they don't upload nonpublic documents to free, or "open," AI tools, while others use AI tools only in private or incognito modes.
The judges also expressed concern that judicial use of AI could undermine public confidence in the courts or degrade the skills of young lawyers and law students.
"One judge described this potential loss of skills when outsourced to GenAI as 'cognitive atrophy.' Another judge worried that all legal writing would start to look the same, as too many lawyers overly relied on GenAI assistance in their work," researchers said.
And the judges worried that the technology could do away with the jobs of court transcribers, clerks and recent law school graduates, who traditionally do work like summarizing depositions or drafting discovery requests, according to the report.
The judges expressed "unease" with GenAI's growing use by self-represented litigants, which could result in a flood of lawsuits, with one interviewed judge describing "dramatic" increases in pro se litigants and their filings.
Researchers also asked the judges about what AI training they think would be helpful. Jurists suggested that they're especially interested in training in prompt engineering and how to spot hallucinations and errors. Judges said they want to learn how other courts are using GenAI, so that they can share best practices, as well.
But all the judges interviewed insisted that GenAI can help — but not replace — judges as human decision-makers.
"Soon, AI usage will be so ubiquitous that it will be difficult to imagine a time when everyone wasn't using it. For judges to have the foresight to learn about these technologies and how it can potentially impact the administration of justice — for good or ill — is an important step," said report co-author and NCSC principal court research associate Jennifer Elek.
--Editing by Kelly Duncan.
Update: This story has been updated with comments from the report's co-authors.
For a reprint of this article, please contact reprints@law360.com.
