A California appeals court has appointed a judge to dig deeper into how errors generated by artificial intelligence made their way into criminal court filings in Nevada County, a case that could result in sanctions against the county’s district attorney.
The move comes after prosecutors in the small Sierra Nevada foothill county, east of Sacramento, repeatedly used generative AI tools last fall in ways that introduced factual mistakes into court documents. District Attorney Jesse Wilson’s office has already acknowledged that four felony cases were affected.
The appointment follows a sworn declaration filed earlier this month by Madison Maxwell, a former Nevada County deputy district attorney who authored several of the flawed filings. Maxwell claimed that when she tried to come forward and correct the mistakes, her supervisor blocked her from immediately notifying the court, defense attorneys and the defendants involved.
Maxwell’s statement was submitted as part of her defense against potential sanctions, which could include monetary penalties or a referral to the State Bar of California for possible misconduct.
Three of the four cases in question were supervised by Maxwell, including one involving a habeas corpus petition seeking release of a man from county jail. All four involved AI-generated content — either drafted or edited using the technology — that introduced falsehoods into official court records.
The newly appointed judge’s review stems from a directive issued by the California Supreme Court in January. The high court asked the appellate court to examine more closely how far-reaching the AI-related errors were and whether they tainted the outcomes of criminal cases. In doing so, the Supreme Court reversed a lower court decision that would have shielded Wilson’s office from any sanctions.
Legal experts say the Nevada County episode is notable because it represents one of the first documented instances nationwide in which government prosecutors — who are typically held to a stricter ethical standard than private attorneys — have been accused of presenting AI-generated falsehoods to a court.
The Nevada County District Attorney’s Office did not respond to a request for comment on Maxwell’s allegations or the appellate court’s decision.
In her declaration, Maxwell said her direct supervisor, Assistant District Attorney Lydia Stuart, had encouraged the use of AI tools in the office without establishing any formal policy or training on how to use them responsibly. She said the office had a culture in which young attorneys were frequently left with little oversight from senior staff.
Maxwell also cited personal health struggles, saying she experienced cognitive difficulties tied to poorly managed diabetes during the period in question. She said a reorganization within the district attorney’s office around that time led to the departure of several experienced attorneys, leaving her with a heavier caseload and fewer mentors to consult.
“Even today, Ms. Maxwell cannot say with certainty which errors were caused by her own drafting errors, as opposed to AI, as opposed to the brain fog and other neurological symptoms brought on by her then-unmanageable diabetes,” her declaration states.
Maxwell also said she was unaware, at the time, that AI programs could generate citations to court cases that do not actually exist. She joined the district attorney’s office in July 2024 — her first job out of law school — and resigned this past May. She now works for the Sutter County District Attorney’s Office.
Wilson and Stuart previously confirmed that four felony cases contained errors consistent with AI use. They said an audit covering roughly 18 months of case filings turned up no broader pattern of AI-related mistakes beyond those four cases, which came to light in February. It remains unclear whether that internal audit has been made public.
More than 20 artificial intelligence experts have urged the California Courts of Appeal to order a more thorough investigation into what happened in Nevada County. Civil Rights Corps, a nonprofit legal organization that assisted the county’s public defender’s office in filing habeas corpus petitions and requests for sanctions, said Maxwell’s declaration only reinforces the need for independent scrutiny.
“An independent investigation is needed to understand the scope of what really happened here and how many people and cases have been impacted,” said Peter Santina, an attorney with Civil Rights Corps, speaking to CalMatters shortly before the appellate court’s ruling.
The California Public Defenders Association also weighed in, filing a brief in support of sanctions. The group argued that prosecutorial misuse of AI is especially concerning because such errors are less likely to be discovered than similar mistakes in civil litigation. Kate Chatfield, the association’s executive director, said the stakes are especially high when a prosecutor cites a nonexistent case to justify keeping someone behind bars.
“That has serious implications for that defendant and their family,” Chatfield said.
The Nevada County case is part of a broader pattern of AI-related problems surfacing in courtrooms across California and the country. Independent researcher Damien Charlotin has tracked more than 130 instances nationwide in which the use of generative AI led to court filings containing fabricated information, including fictitious case citations. His database includes a recent lawsuit involving State Farm over a fire-related insurance dispute.
Last year, a panel of judges issued a $10,000 fine against a Los Angeles-area attorney who used ChatGPT to draft a filing containing more than 20 fabricated quotations — a sanction widely viewed as a warning to attorneys across the state. Since then, according to Charlotin’s tracking, 19 attorneys and self-represented litigants in California have faced monetary sanctions tied to AI-related errors in their filings.
Even as courts crack down on AI misuse by attorneys, judges and court staff themselves have begun experimenting with the technology. Pilot programs testing AI tools are already underway in Los Angeles and Riverside counties, drawing scrutiny from legal observers. Separately, federal courts have also seen instances of judges’ rulings containing AI-related errors.
Technology experts at UCLA School of Law have told CalMatters that overworked attorneys, judges and court staff are often the ones most likely to turn to AI tools for help — and the least likely to have the time to verify that the technology hasn’t fabricated facts or invented case law along the way.
Original source: CalMatters
