An Oklahoma attorney just got a very public lesson in AI reality-checking. The Oklahoma Supreme Court publicly reprimanded lawyer Matthew B. Reeves on May 27, 2026, after finding he filed motions packed with case citations generated by ChatGPT that he never independently verified. The justices chose a formal public reprimand instead of a suspension, but the signal was clear: judges expect lawyers to treat AI output as a starting draft, not the final word. The ruling joins a growing line of decisions holding attorneys responsible when generative tools cough up fake authorities.
In its written decision, the court said Reeves admitted he had used ChatGPT to obtain five case citations that he dropped into two motions, then "failed to verify the ChatGPT citations through independent review," according to the Oklahoma Supreme Court. In disciplinary docket 2026 OK 37, the court entered a public reprimand and noted the sanction tracked similar penalties Reeves had already received in another jurisdiction.
The rebuke traces back to a July 23, 2025, federal sanctions order in Alabama that found several of the cited cases simply did not exist and described the bogus authorities as "hallucinations" produced by ChatGPT, according to the federal court. National coverage by AP reported that the episode helped push judges and ethics panels to treat AI-generated citations as squarely the attorney's responsibility.
The matter drew renewed attention in Oklahoma after the attorney general flagged alleged AI errors in a 132-page motion filed in the fraud case involving barbecue owner Brent Swadley. The Swadley defense team denied using AI in that filing, as reported by News 9. The flap prompted bar leaders and local judges to circulate fresh reminders about verification and training for lawyers who rely on generative tools.
What the Court Found
Writing separately, Justice Gurich did not mince words, warning that "the unverified use of generative artificial intelligence resulting in reliance upon false authority is reprehensible and clearly violates numerous ethical rules." The court concluded that Reeves's conduct breached multiple Oklahoma Rules of Professional Conduct, tying the misconduct to duties of diligence, candor to the tribunal, and the prohibition on conduct prejudicial to the administration of justice. It imposed a public reprimand; see the Oklahoma Supreme Court analysis for the full breakdown.
Legal Implications for Lawyers
The Oklahoma Bar Association has been urging caution for months and has circulated a tip sheet telling lawyers to treat AI outputs like any other untested draft: verify every citation against primary sources, protect client confidences, and adopt clear internal rules on when and how AI can be used. The Oklahoma Bar Association frames those steps as straightforward applications of existing competence, candor, and confidentiality requirements, not brand-new obligations.
Why This Matters in Oklahoma
State courts are already hard-coding those expectations into their rules. In February, the Oklahoma Court of Criminal Appeals adopted a rule that requires attorneys to verify any portion of a filing that was produced or altered by generative AI, and local judges are weighing whether similar language should show up in district court dockets. The order and its ripple effects have been closely watched across the state bar.
For litigators and law firms, the takeaway is straightforward: AI can speed up drafting, but it cannot replace the human work of checking cases and facts. The Reeves reprimand shows that skipping that step does not just risk embarrassment, it can lead to public discipline that trails an attorney across jurisdictions.
