US Courts Cautiously Experiment With AI to Speed Up Their Work

Federal courts are warming up to artificial intelligence.

Judges, clerks, and other staff are experimenting with the platforms for routine administrative tasks. The federal judiciary has a task force to examine whether courts need new policies on the use of AI.

While courts are still testing AI, judiciary employees are cautiously optimistic that it could help courts be more efficient—and potentially cut some costs along the way.

Former US District Judge Paul Grimm said courts became wary of AI after a few high-profile incidents of lawyers including “hallucinated” non-existent cases in their briefs that were generated by the platforms.

He said tools have become more reliable, and the judiciary could generate more public confidence by using them—as long as they’re properly and seriously tested.

“It has the potential to enhance the reputation of the legal system,” Grimm said.

Testing AI

For many courts, use of AI is still being debated.

Lyle Cayce, the clerk for the US Court of Appeals for the Fifth Circuit, offers an “attorney toolbox” for lawyers, meant to help them comply with the rules for filing briefs. The program, which doesn’t use AI, alerts them if a filing may not comply with the rules.

He said they’re now testing whether AI could run the same functions. Cayce said it wouldn’t be able to work more quickly than his current program, but it could handle more briefs at a time.

Cayce said they’re also looking at using AI to make initial evaluations on motions for extensions of time. “Obviously a human’s going to make that decision, but AI can actually speed that up quite a bit,” Cayce said.

He said they’re also experimenting with using AI to determine whether a party filed an appeal in time for it to be considered. And Cayce said AI could be used to help process pro se filings, which can be handwritten and sometimes harder to understand.

Cayce said he plans to talk about his office’s experiments at a meeting of appellate clerks toward the end of April, and expects to have integrated at least one or two uses of AI into their workflow by then.

Cayce said he’s staying away from using AI to evaluate legal arguments. “I’m more interested in using narrow AI to either simplify court processes or to identify issues that could be resolved later,” he said.

Susan Goldberg, the circuit executive for the First Circuit, said Chief Judge David Barron has appointed a working group on the issue. She said that group has held a number of discussion forums since September of last year, with topics from how clerks’ offices could use AI to how courts are addressing it in their standing orders.

The Federal Circuit is also beginning to test out AI. Jarrett Perlow, the circuit’s clerk, said while the court isn’t doing formal testing, he’s tinkering with some of the tools “just to see how it might be able to enhance some of the work that we’re doing here.”

Perlow said an AI platform could generate a report on court statistics in seconds while it might take a court staffer hours to produce the same work.

He also said any use of AI by the circuit would be done cautiously.

Court administrators said it’s vital to not put confidential documents on platforms that could be accessed by others, and that they’ve only tested the systems by using publicly available information.

Judicial Use

US District Judge Xavier Rodriguez in San Antonio handles all of his initial status conferences rather than referring them to a magistrate judge. He said, in order to keep up with his work, he’ll use an AI tool to review the publicly filings and generate summaries and a timeline of events.

“Then I can have, I think, a more meaningful status conference with lawyers, especially when they’re having issues about how much discovery should be taken,” Rodriguez said.

Rodriguez said members of his court, the Western District of Texas, are also talking about potentially using AI to check that lawyers’ briefs comply with the rules, help members of the public understand what stage of litigation their case is in, and to produce internal data on things like judges’ workloads.

Rodriguez said that while he uses AI to help prepare for those conferences, he’s more hesitant of using it to write orders. He said judges should understand that the results they’ll get from any platform depends on how that model was built and trained.

AI Task Force

The judiciary’s AI task force is meant to serve as a “coordination hub” for policy issues created by the technology, according to a description viewed by Bloomberg Law. The task force, meant to complete its work by the end of 2026, “will determine the need to amend or establish policies on the judiciary’s use of AI.”

The group—made up of judges and other court staff—will make recommendations for policy making groups within the judiciary and the Administrative Office for the US Courts.

The AO has also created an “AI governance framework and job aid to help advise judiciary leaders on AI-related topics,” according to a report of the Judicial Conference’s September meeting.

Rodriguez is one of several judges who worked on a separate AI working group for the American Bar Association, which published recommendations in February for potential usage by courts.

He said those guidelines are meant to give courts a “yellow light” to use the tools—"proceed with caution.”

Rodriguez is not serving on the judiciary’s AI task force. But he said he hopes the group breaks down its evaluation of the technology by different uses.

“Using AI tools for back-end purposes, finances, admin, human resources is one thing,” Rodriguez said. “How you use it for judicial decision making is another thing.”