Judicial Council task force to weigh use of AI in courts

AI might be an exciting technology for courts in the future, but for the moment it poses risks for privacy and bogus citations. At least, that was the message from a report presented to the Judicial Council on Friday.

“We need to have a little talk about generative AI,” said Mary J. Greenwood, administrative presiding justice of the 6th District Court of Appeal.

Greenwood was introducing a report on “Generative Artificial Intelligence and California’s Judicial Branch.” During the Judicial Council’s meeting in January, Chief Justice of California Patricia Guerrero asked Greenwood and Alameda County Superior Court Judge Arturo Castro to lead an effort to research the potential use and pitfalls of AI in the courts.

After acknowledging that neither of them claimed to be experts on the topics, Greenwood noted three clear dangers. The first was illustrated by an attorney who was sanctioned last year.

“On a day of infamy in the New York federal court, an unfortunate member of the bar submitted a brief that contained citations that didn’t exist,” Greenwood said. “I can only imagine what would have happened had he been appearing in front of [California Supreme Court] Justice [Carol A.] Corrigan under those circumstances.”

She added that the incident happened because the attorney didn’t understand the “generative” part of the software he was using. He thought he was using a “super search engine” that “found cases that weren’t in LexisNexis.” ChatGPT responded to his prompts by creating citations that “sounded authoritative” but didn’t exist.

“He had practiced law for 30 years,” Greenwood said. “I think that was probably part of the problem… He simply didn’t understand the ramifications of what he was doing. His kids were in college and were using ChatGPT.”

Some courts, including the U.S. Central District of California, now require attorneys to disclose whenever they use AI in writing briefs. LexisNexis maintains an online list of courts that require disclosure.

Greenwood and other speakers also noted that AI can create privacy concerns for court users. If AI systems are allowed to search through millions of pages of court documents, they are likely to come across information that should have been redacted but wasn’t, including Social Security numbers.

Finally, Greenwood said if judges use AI in their work, they risk “improper delegation of judicial decision-making.” From the perspective of a judge, she said the most important thing to remember is that judicial ethics always apply.

“I can almost hear [ethics specialist] David Rothman in my ear right now saying, ‘Number one, you are always a judge,’” Greenwood said. “No matter where you are, you are always a judge. It’s clear that some canons will always apply.”

Senate Judiciary Committee Chairman Tom Umberg, D-Santa Ana, warned the council that while the technology races ahead, time is running short if it needs legislation passed this year regulating the use of AI in courts. He noted that all bills need to be on Gov. Gavin Newsom’s desk in the next 104 days. He said he has heard of AI-based software that can write memos suggesting legal decisions or create “artificially enhanced evidence.”

“Does the Legislature step in and further define spoliation of evidence, further define standards, or will the judiciary do that?” Umberg asked. “We have some real challenges.”

After the presentation, the council voted to create an Artificial Intelligence Task Force to oversee guidelines for AI in the courts. It also directed the California Supreme Court’s ethics committees “to consider amendments to the Code of Judicial Ethics” and the Center for Judicial Education & Research to create educational materials.

Guerrero closed by officially announcing the creation of the task force, saying it will be led by 5th District Court of Appeal Presiding Justice Brad R. Hill. It will also include Justice Greenwood, Judge Castro and the leaders of several policy committee chairs within the council.

Guerrero closed by noting that Google, WestLaw and LexisNexis all already use AI tools. She added that “so much has changed” in the year since she addressed graduating students at UC Davis School of Law.

“I told them that their jobs were safe,” Guerrero said. “I hope I’m not proven wrong.”