Hawaiʻi Circuit Judge Jordon Kimura wasn’t persuaded by Mark Valencia’s explanation of why he submitted a brief citing a non-existent case and misstating a Hawaiʻi Supreme Court decision to support his request for the judge to toss the case against his client.

Valencia blamed it on his associate, Christilei Hessler, who he said had used an artificial intelligence program to draft the brief and he didn’t bother to verify the citations before filing the brief. The citations were AI-generated hallucinations, Valencia wrote.

It was, he said, an honest mistake.

Unmoved, Kimura, who had discovered the errors while reviewing the brief, found that Valencia had violated a civil procedure rule requiring that representations to the court be accurate. The judge said he would decide how much to sanction Valencia when the ongoing case ends.

Valencia did not respond to a request for comment.

The incident underscores a growing problem in Hawaiʻi courts, as the use of artificial intelligence takes hold in the practice of law. On Monday, a panel of judges and other legal experts presented a long-awaited report on AI and the court system to acting Chief Justice Sabrina McKenna.

The report explores numerous ways that legal professionals, including lawyers, judges and court administrators, can use AI tools, albeit with training and human oversight to reduce risks, to do things like process documents, create court transcripts and minutes, and provide helpful information to non-lawyers representing themselves in litigation, as well as people who don’t speak English.

The report also discusses ethical issues, including the prohibition against making false statements to judges and opposing counsel, such as the ones Valencia made.

“If attorneys fail to exercise independent judgment and fail to validate and verify outputs from AI tools,” the report says, “these are the potential adverse effects: false and inaccurate contentions are presented to the courts, causing chaos, waste, inefficiency, distrust, and failed advocacy.”

While the report discusses a number of court rules implicated by lawyers using AI, the central one is the rule Valencia was found to have violated, known as Rule 11 of the Hawaiʻi Rules of Civil Procedure. It essentially requires, among other things, that statements made to the court be truthful. And it allows judges to impose sanctions for violations.

The report concludes that Rule 11 is “broad enough to provide adequate safeguards.”

In practice, the safeguards haven’t prevented lawyers from submitting AI-generated fake citations to Hawaiʻi courts. And some lawyers have faced no penalty for doing so.

Response To Fake Cases Varies Widely

Around the same time Valencia was explaining to Kimura why he had submitted a brief with fake, AI-generated citations, another lawyer from his firm, Case Lombardi, was doing the same thing before another judge.

In that case, Kaʻōnohiokalā J. Aukai IV had submitted a brief in which all six of the document’s citations were flawed: two of them completely fabricated, the others misstating what real cases actually said. All, he explained to the court in an apology, were the result of AI hallucinations.

In Aukai’s case, Circuit Judge Kelsey Kawano accepted Aukai’s apology and opted against sanctions. In the case of Honolulu Deputy Corporation Counsel David Sgan, who used AI-generated fake law in a brief filed in a major civil rights case against the Honolulu Police Department, Circuit Judge Karen Holma let the lawyer off with a scolding.

Hawaiʻi federal courts, by contrast, have taken a hard line. Lawyers must tell the court when they use AI to produce any court document and verify that they’ve confirmed material in the document isn’t fake. Lawyers who submit fictitious material face sanctions under Federal Rules of Civil Procedure, the order indicates.

The Hawaiʻi Supreme Court is considering doing the same, the report says.

A Growing, Worldwide Problem

Hawaiʻi courts are hardly the only ones where AI is leading lawyers and judges astray. French researcher Damien Charlotin has a growing database of legal decisions addressing lawyers using generative AI to produce hallucinated citations and other ersatz material. Charlotin had documented 690 cases worldwide as of Monday, including 470 in the U.S., more than double the number in late August.

As in Hawaiʻi, lawyers are rarely sanctioned. One notable exception involved a case brought against the Chicago Housing Authority. A jury had hit the authority with a $24 million verdict after finding the plaintiffs’ children had suffered permanent brain damage from living for years in an apartment owned by the authority.

“Officers of the court cannot become comfortable with careless or deliberate misrepresentation of facts or the law.”

Cook County, Illinois, Circuit Judge Thomas Cushing

After the trial, the housing authority’s lawyers filed a lengthy motion arguing the court erred in prohibiting the authority from presenting certain evidence to the jury. After the authority’s lawyer admitted that the firm had used AI to generate a single flawed citation in the brief, opposing counsel pointed out 14 such flaws. 

The court imposed sanctions of $59,500 against the authority’s firm to reimburse opposing lawyers for the time they wasted trying to research the flawed citations. 

The court noted that the practice of submitting false citations and factual claims poses “a grave threat to the judicial branch.”

“The authority of the courts relies on public confidence that rulings are just and are grounded in the law, not on the whims of judges,” the court said. “Officers of the court cannot become comfortable with careless or deliberate misrepresentation of facts or the law.”

What sanctions Kimura imposes on Valencia remains to be seen. Although Valencia blamed his associate for the flawed brief and noted that it was the judge and not opposing counsel Bosko Petricevic who had discovered the errors, Valencia acknowledged he was ultimately responsible.

In the future, he told the court, he’ll make associates include cited cases along with the briefs submitted to him for review.

“This kind of mistake will not happen again,” he wrote.