Lawyer Faces Fresh Reprimand Over AI ‘Hallucination’ Filings as Fallout Widens

A former law firm partner who was sanctioned by a federal judge in Alabama last year over artificial intelligence-generated errors in court filings has been reprimanded again by the Oklahoma Supreme Court, marking the fourth time he has faced formal disciplinary action over the same incident.

“Human diligence and review is required to ensure content and accuracy of filed documents,” the Oklahoma Supreme Court said in its order. “Signing pleadings that contain citations generated by artificial intelligence that have not been verified shows a reckless disregard for the truth and an indifference to accuracy.”

The lawyer, Matthew Reeves, was a partner at Butler Snow last year when he was sanctioned for fabricated citations in two court filings in a case in which the firm was representing the Alabama state Department of Corrections.

Reeves, who is no longer listed as a Butler Snow attorney on the firm’s website, had urged the state justices not to impose any further discipline, citing “the reputational, professional and personal consequences I have already faced and will continue to face”.

Reeves and Butler Snow did not immediately respond to requests for comment.

The adoption of generative artificial intelligence by lawyers has led to increasing cases of the technology producing non-existent legal sources, misquoting the law or fabricating case citations in court filings, prompting courts to sanction attorneys who fail to properly vet AI-assisted submissions.

The Oklahoma order, which adds to prior sanctions against Reeves for the same AI-related errors, underscores the risks lawyers face when failing to verify AI-generated material. In addition to Oklahoma and Alabama, Reeves has been publicly censured in Texas and Tennessee under rules allowing courts to discipline attorneys licensed in their states if they have been sanctioned elsewhere.

In a filing in December, Reeves said he accepted responsibility for failing to verify five case citations he said were generated by OpenAI’s ChatGPT.

“I did not fully appreciate the specific problems associated with AI use in legal research, including the risk of ChatGPT producing valid legal propositions with fabricated, non-existent citations,” Reeves told the court.

OpenAI did not immediately respond to a request for comment.

Reeves said in the filing that he has teamed up with University of Alabama law professor Anil Mujumdar “to develop an informative programme to educate law students about the risks of AI”.

Mujumdar, a law professor at the university, did not immediately respond to a request for comment. He was Reeves’ opposing counsel in the Alabama case in which the AI errors occurred.

Butler Snow, which employs more than 300 lawyers, said in response to last year’s sanctions that it would conduct “additional and extensive firm-wide training on the appropriate use of artificial intelligence”.

For any enquiries or information, contact ask@tlr.ae or call us on +971 52 644 3004. Follow The Law Reporters on WhatsApp Channels.