US judges form group to tackle pitfalls and promise of AI

By Sara Merken

NEW YORK, Jan 26 (Reuters) - As artificial intelligence steadily seeps into U.S. courtrooms, a group of state and federal judges has joined forces to confront the technology head‑on.

The Judicial AI Consortium, launched last week by U.S. Magistrate Judge Maritza Dominguez Braswell in Colorado, U.S. District Judge Xavier Rodriguez in Texas and Judge Scott Schlegel of Louisiana’s Fifth Circuit Court of Appeal, aims to connect judges nationwide to swap ideas, discuss emerging risks and share how they are already using AI on the bench.

“GenAI is here, and we have to start thinking through proper use cases,” Schlegel said.

About 90 judges have already expressed interest, Braswell said. The goal is not to develop blanket policies but to share information and resources so that judges can make better decisions for themselves on how to handle AI-related issues, she said.

The spread of generative AI over the past three years has forced courts across the country to grapple with its place in legal practice. AI chatbots can shave precious time off legal research and drafting, but they can also generate fictional details or made-up or misidentified case citations, known as "hallucinations." That has led to sanctions, stiff fines and pointed warnings from judges about the need for lawyers to verify their work.

Two federal judges disclosed in October that their staff had used AI to help draft court orders, resulting in errors that later drew scrutiny from Congress. U.S. Senate Judiciary Committee Chairman Chuck Grassley said on the Senate floor that the flawed rulings should “serve as a wake‑up call across the federal judiciary.”

Several states have moved to adopt AI policies for judges and court personnel, acknowledging both the efficiency the tools can offer and the dangers of overreliance.

Rodriguez said he expects judges during the group’s first meeting, set for next month, to raise a range of topics for the group to discuss -- from practical applications of AI in chambers to thorny issues like AI hallucinations and deepfake evidence.

There’s a sense among some judges that there’s an information "vacuum" -- people don’t know what tools are out there or how they should be used, Rodriguez said.