A New Mexico attorney representing a convicted murderer submitted an appellate brief containing fabricated police statements and witness accounts after using ChatGPT to condense the trial record, according to the New Mexico Supreme Court.

The court found that attorney Stephen Aarons failed to verify the material generated by the artificial-intelligence system. The judges said the brief “contained false testimony from wholly fabricated witnesses.” On Wednesday, the court held Aarons in contempt and ordered him to pay $5,000. His conduct will also be referred to an attorney disciplinary board.

The judges said Aarons had “demonstrated a lack of remorse and a lack of concern for his client.”

Aarons, a private attorney based in Santa Fe, said he used ChatGPT last year after taking over the appeal to condense the trial record. He said he did not understand how extensively AI “hallucinations” could occur or how confidently software could generate facts about events that never happened.

“I am remorseful but hopeful that the disciplinary board takes into account it was an honest mistake,” Aarons told reporters, according to the report. “It is a lesson learned for all professionals who rely upon this powerful but sometimes unstable technology.”

Aarons represented Oscar Renee Sandoval, who pleaded not guilty to killing the mother of his children. Sandoval was later convicted and sentenced to life in prison last year. His appeal remains pending. The contempt order concerns the accuracy of Aarons’ filing, while the underlying appeal is now being handled by different counsel.

Sandoval’s case was transferred to Kim Chavez Cook, a New Mexico public defender, on September 2. Cook is now handling the appeal while the case remains open.

Lawyers in both state and federal courts have previously faced sanctions over the uncontrolled use of AI. In several cases, attorneys submitted briefs containing fabricated cases, false references or inaccurately quoted legal provisions. Aarons’ case is distinct because the filing included false testimony attributed to entirely fabricated witnesses in a criminal appeal.

The case was reported by Cryptopolitan, which also linked to a Reuters report.

Florida lawsuit alleges ChatGPT chats were connected to FSU shooting

A separate legal dispute concerns ChatGPT and the shooting at Florida State University. An attorney representing the family of Tiru Chabba, who died in the shooting, has sued OpenAI, ChatGPT and Phoenix Ikner in federal court.

The lawsuit was filed Sunday in the United States District Court for the Northern District of Florida. It alleges that OpenAI should have taken action in response to the activity described in the case. The filing presents the family's claims; it does not by itself resolve whether OpenAI is legally responsible.

Florida law enforcement officials released transcripts of Phoenix Ikner’s ChatGPT conversations. The transcripts include queries about operating a gun, media coverage, the busiest times at FSU, the Columbine shooting and how to become infamous.

Bakari Sellers, an attorney for Chabba’s family, said the exchanges extended over approximately 18 months and totaled about 16,000 “disturbing chats.”

“This is the same person who asked, you know, how can he become infamous? He asked about the Columbine shooting. He asked about what time should he go to campus? What time are most people going to be there?” Sellers said.

“He literally utilized open AI and Chat GPT as his co-conspirator, utilized it as a resource to carry out mass murder,” Sellers added. “There was nothing in place to prevent that from happening and so lives were lost. That’s the inherent danger, there has to be something in place to prevent that from happening.”

OpenAI rejected responsibility for the shooting. Spokesperson Drew Pusateri said the company located an account believed to be associated with Phoenix after learning what had happened and proactively shared the information with police. Pusateri said OpenAI continues to cooperate with authorities.

“Last year’s mass shooting at Florida State University was a tragedy, but ChatGPT is not responsible for this terrible crime,” Pusateri said.

He said ChatGPT provided factual material that was already widely available through public internet sources and did not encourage illegal or harmful conduct. Pusateri also said hundreds of millions of people use ChatGPT every day for legitimate purposes.

According to Pusateri, OpenAI is continually improving tools designed to identify user intent, prevent misuse and respond to potential safety concerns.

The lawsuit followed a similar case in Florida. Attorney General James Uthmeier recently disclosed that the Office of Statewide Prosecution had opened a criminal investigation into OpenAI and ChatGPT.