News
Generative AI Hallucination in Court Filing May Jeopardize Case Against BART
Comments
Link successfully copied
The screen shows the ChatGPT app in this photo illustration. (Oleksii Pydsosonnii/The Epoch Times)
By Lear Zhou
5/25/2026Updated: 5/25/2026

SAN FRANCISCO—A California attorney was sanctioned $1,000 for a court filing that refers to three nonexistent cases, indicating possible use of artificial intelligence that hallucinated, an issue increasingly seen in courts.

“The Court does not prohibit or oppose the use of artificial intelligence in legal advocacy, so long as counsel independently verifies the accuracy of AI-generated content,” Magistrate Judge Thomas S. Hixson wrote in the order on May 6.

The sanctioned attorney, Jessica R. Barsotti, represents the plaintiff, Karl Gregoire, in a case against Bay Area Rapid Transit for firing a fare inspection officer who, during the COVID-19 pandemic, rejected the mandated COVID-19 vaccine on religious grounds.

Barsotti filed the questionable court filing in response to the defendant’s motion for sanctions after Barsotti had been absent and missed a few deadlines. The requested sanctions included dismissal, prohibiting the plaintiff from presenting evidence, and a monetary penalty.

Barsotti said in the file that she has experienced a “relapse of pre-existing medical condition” following a series of stressful events, including separation from her husband, her law firm partner’s retirement, and the death of a close family member.

She said she ended up in a treatment facility for 75 days to address the medical and emotional crisis and was unable to attend court meetings or respond to any actions needed.

Barsotti cited in this document four cases arguing that the plaintiff’s interest should not be affected by the counselor’s failures; three were found by the court later to be nonexistent.

The attorney’s errors may jeopardize the lawsuit; the judge has issued a sanction warning of “striking the filing in whole or in part.”

Other sanctions against Barsotti include a request that she inform the plaintiff of the sanction and attend at least one hour of education on the “ethical use of artificial intelligence in the practice of law.”

The judge ordered the clerk to inform the State Bar of California, a regulatory agency overseeing California attorney licensing and discipline, about the incident.

The Epoch Times was unable to reach Barsotti for comment.

More than 1,000 U.S. legal cases in which generic AI produced hallucinated content—typically fake citations, false quotes, misrepresentations, and outdated advice—were detected from the second quarter of 2023 to May 19, which account for nearly 70 percent of such cases globally, according to researcher Damien Charlotin.

The capabilities of generative AI in the legal field have been noted since March 2023, when the newly launched GPT-4 model by OpenAI surpassed 90 percent of humans in passing the bar exam.

According to a 2025 survey by the International Legal Technology Association of 580 law firms worldwide—representing more than 152,000 lawyers—88 percent of participants believe that AI, in some form, will create significant change in the legal technology profession in the next three to five years.

“The use of artificial intelligence is a recurring problem in the legal field,” Kevin Snider, chief counsel for the Pacific Justice Institute, told The Epoch Times via email.

“AI will give a nonexistent case, or it will give a real case that gives a false legal holding.

“If false AI-generated evidence is known to be used in court proceedings, public confidence in the legal system will collapse.”

Snider was the lead attorney representing six former Bay Area Rapid Transit employees and won $7.8 million in damages for them after they were fired or forced into retirement for declining the COVID-19 vaccine in October 2024.

The State Bar of California has filed notice of discipline charges against five attorneys for potential misuse of AI, with two cases approved by the State Bar Court and awaiting final decisions by the California Supreme Court.

No attorneys in California have been disciplined for filing AI-generated content so far.

“These cases underscore that failing to review and confirm the accuracy of one’s submissions—regardless of the tools used to create them—can cause real harm and undermine the administration of justice,” George Cardona, the state bar chief trial counsel, said in an email statement to The Epoch Times.

Share This Article: