Civil Rights Corps Asks Court of Appeal To Investigate a California Prosecutor’s Office for Repeated Use of Fabricated, Seemingly AI-Generated Legal Citations

“Even as AI-hallucinated legal authority plagues courts across the country, the Nevada County District Attorney’s Office’s actions appear to be unprecedented”

Nevada County, CA – On March 30, 2026, Civil Rights Corps (CRC) filed a brief in support of sanctions against a District Attorney’s Office in the case of Kjoller v. Superior Court of Nevada County, urging the court to conduct an inquiry into a series of filings by Nevada County prosecutors riddled with citations to fabricated legal authorities, fictitious quotations, and misrepresentations of law. Below is a background summary and quotations from that brief.

The case started with a petition for writ of habeas corpus that CRC and the Nevada County Public Defender’s Office filed on behalf of Kyle Kjoller, a man who the Court of Appeal ultimately agreed was unlawfully incarcerated pretrial on an unconstitutional detention order. The District Attorney’s Office filed a response with fabricated and nonexistent legal citations, so CRC filed a Motion for Sanctions and asked the court to investigate. The court denied the motion. Soon after, CRC found that the same prosecutor’s office had included serious citation errors in three additional briefs in three separate cases, including non-existent quotations. Then prosecutors filed two “amended” briefs that left serious citation errors remaining. 

CRC filed a Petition for Review in the California Supreme Court. Scholars and public defenders filed amicus letters urging the court to investigate. Scholars explained that a court-run investigation is necessary for uncovering the root cause of the apparent AI errors and for mitigating the risks of prosecutorial misconduct and wrongful convictions; public defenders across the state explained how the power and immunity of prosecutors and the chaotic pace of busy courts make citations to fabricated authorities especially dangerous for indigent clients.

In January 2026, the California Supreme Court ordered the Court of Appeal to reopen the sanctions matter and issue an Order to Show Cause directing the District Attorney’s Office to explain why sanctions should not be imposed. The Court of Appeal issued the Order and will decide whether to appoint a referee to conduct evidentiary hearings on these issues.

The District Attorney’s Office filed its Response on February 27, 2026, conceding that it committed sanctionable misconduct and that its errored citations showed the hallmarks of AI, arguing that the court should decline to take further action because the Office had handled the issue internally. The District Attorney’s Office said the prosecutor who submitted several of the briefs was taken off her cases and the Office was investigating.

On March 30, 2026, CRC filed a Reply in Support of Motion for Sanctions. Below are quotations from that brief (with internal citations removed):

• “For months, the Office appears to have misled courts, counsel, and the public. Prosecutors expressed outrage at the idea that AI hallucinations were present in their briefs; the second-in-command of the Office even told opposing counsel that such an accusation was unethical and threatened consequences if counsel persisted in moving for sanctions.” (page 8)
• “Even after being notified of the fabricated authority, prosecutors allowed briefs with erroneous citations to linger on court dockets for weeks without promptly correcting them.” (page 9)
• “The Office now admits that its citations bear ‘all the markings’ of AI hallucinations and that the prosecutor who drafted many of the briefs is being investigated. It further claims that it has addressed the issue internally by instituting unspecified policy changes as well as warning and training its staff about the dangers of AI use. This claim is not new; the Office told the superior court it had resolved this issue by taking these exact same steps, even while briefs with fabricated citations sat in court dockets.” (page 9)
• “The Office claims its training and AI policies–the same described to the superior court in November–render sanctions unnecessary. But none of those trainings or policies (which were not submitted to the Court and counsel has not seen) compelled the Office to stop filing briefs citing to fabricated authority, notify courts of the fabricated authority which had previously been filed, or retract its misstatements to courts or opposing counsel.” (page 35)
• “Reliance on fabricated citations is especially dangerous when the rights of a criminal defendant are at stake. But in this case, the Office repeatedly asked courts to rely on made-up caselaw to deprive indigent criminal defendants of their most basic constitutional rights. If, as seems likely, the errors were actually caused by AI hallucinations, prosecutors lied about the source of the errors to courts, opposing counsel, and the public.” (page 48)
• “[S]erious and essential questions remain open. Not about whether this conduct deserves to be sanctioned–courts routinely sanction lawyers for far less–but about what actually happened at the Office that led to fabricated authorities being cited in so many cases, management’s response, and how many other briefs with errored citations remain outstanding. This Court still does not know:
  1. What is the source of these citations to fabricated authorities? The Office still refuses to squarely address the source of the fabrications; the Office has not confirmed that its citations to fabricated authorities are the product of the reckless inclusion of AI hallucinations. The OSC Response does not even include a declaration from the prosecutor who filed the briefs.
  2. What supervision regarding the use of AI and citation-checking were prosecutors provided during this time period? When did high-ranking supervisors first learn of, or suspect, that AI hallucinations were present in briefs? For months, the Office’s leadership made a series of apparently false statements and claimed opposing counsel acted unethically for suggesting their briefs may contain AI hallucinations. How much did the supervising lawyers know at the time?
  3. How many other briefs has the Office filed with similar errors? How many remain pending in California courts? The OSC Response reveals for the first time that the Office has reviewed 18 months of its own briefs and conceded that it found citation errors. To counsel’s knowledge, these errors have not been revealed to this Court, the superior court, or the defense bar. The Office also gives no explanation for how it conducted the audit or categorized those errors as “minor or non-substantive citation issues” as opposed to representatives of a “broader use of errors due to artificial intelligence or any other systemic pattern of error.” The Office offers no explanation for why its current assurances that AI was not used in these cases are more reliable than similar past assurances that now seem to be false.” (pages 10-11)
• “Even as AI-hallucinated legal authority plagues courts across the country, the Nevada County District Attorney’s Office’s actions appear to be unprecedented.” (page 48)

###

Civil Rights Corps challenges systemic injustice in the U.S. legal system. Through innovative civil rights litigation, advocacy, and public education, we aim to re-sensitize the legal system and our culture to the injustice and brutality that characterizes the contemporary legal system. Our work is guided by a commitment to the people and communities harmed by policing, surveillance, incarceration, discrimination, and the criminalization of poverty.