Stephens County judge reportedly used ChatGPT for legal research. Investigators said at least two cases cited in a child custody order were nonexistent, raising concerns about artificial intelligence and judicial accountability.
A judge's courtroom is supposed to be a place where facts are examined, evidence is considered and the law is carefully applied. In Stephens County, a judicial misconduct investigation has placed those basic principles in the spotlight after artificial intelligence was reportedly used during legal research for a child custody case and at least two of the cases cited in the resulting court order were found to be nonexistent.
Stephens County District Judge Lawrence Wheeler reportedly acknowledged to investigators that he used ChatGPT while conducting legal research for a court order. According to an August letter from Stephens County District Attorney Jason HickstotheOklahomaAttorney General'sOffice, Wheeler told Oklahoma State Bureau of Investigation agents that ChatGPT provided him with case citations that were later determined to be nonexistent.
At least two of the cases cited in the order reportedly did not exist.
The order involved a parent's request for psychological testing in a child custody dispute. Wheeler denied the request and cited Oklahoma cases as legal authority for his decision.
Investigators were later told that the cited cases could not be located because they were not real. The order was subsequently vacated and Wheeler is no longer involved in that particular case.
The incident has become an important example of the risks associated with artificial intelligence entering the legal system. The problem was not simply that an AI programproducedinaccurate information.
The greater concern is that unverified information generatedbytheprogramappeared in an official judicial document..
Artificial intelligence has moved rapidly into offices, classrooms, businesses and professional fields, including the legal profession. Attorneys, judges and other professionals can use AI to organize information, locate material, summarize documents and assist with research.
The technology, however, does not guarantee accuracy. AI systems can produce answers that appear authoritative while containing completely fabricated information.
In legal settings, those fabricated responses are commonly referred to as AI 'hallucinations.' They can include nonexistent cases, inaccurate quotations, fabricated citations and incorrect interpretations of statutes.
The Wheeler matter demonstrates the danger of allowing such information to move from a computer screen into a court order without independent verification. It cannot replace the responsibility of a judge to determine whether the information is accurate.
According to information provided to investigators, Wheeler did not simply ask ChatGPT to write the entire order and submit it without review.
He reportedly told investigators thathewrotetheorder himself and used ChatGPT for research. That distinction is significant but it does not eliminate the responsibility attached to the final document.
The citations appeared in an order issued by a judge. A judge's signature carries legal authority. A court order can affect custody arrangements, parental rights, finances, liberty and other aspects of a person's life.
The responsibility for the accuracy of that order rests with the person issuing it, not with the computer program that may have supplied some of the information. When artificial intelligence enters the legal process, the human review process becomes even more important, one might wind up with information not intended like, I took an arrow to the heart,I never kissed a mouth that taste like yours,strawberries and something more .
Someone must verify that thecitedcaseexists.Someone must examine the actual opinion. Someone must confirm that the statute says what the document claims it says.
Someonemustensurethat the facts support the legal conclusion. In a courtroom, that responsibility ultimately belongs to the judge.
The Oklahoma Code of Judicial Conduct requires judges to perform their duties competently and diligently, with the legal knowledge, skill, thoroughness and preparation necessary to fulfill the responsibilities of judicial office. Those standards existed long before artificial intelligence became widely available.
AI creates another avenue through which inaccurate information can enter the legal process but the underlying responsibility remains the same. A judge who receives incorrect information from another person is still expected to verify it before relying on it in an official ruling.
Thesameprincipleapplies when the information comes from an artificial intelligence system. The technology may be new. The obligation to verify the law is not.
AI-generated legal citations can be particularly difficult to identify because fabricated cases can contain realistic names, citation numbers and legal language. Theresultcanlooklegitimate tosomeonewhodoesnotindependently check the source.
That makes verification an essential part of responsible AI use in the legal profession. The Wheeler investigation also highlights the importance of distinguishing between different levels of misconduct.
There is a substantial difference between a judge who uses AI as a research tool, independently verifies the information, discovers an error and corrects it and a judge who repeatedly relies on fabricated information without verification. A firsttime mistake could potentially result in additional training in legal research and responsible AI use.
A serious failure that affects someone's legal rights could warrant stronger disciplinary action. Repeated misconduct, refusal to verify information or intentional deception could lead to substantially more serious consequences under applicable disciplinary rules.
The appropriate response should be based on the conduct, the circumstances surrounding it and the effect on the people involved, rather than simply the fact that artificial intelligence was used. The Wheeler investigation comes as artificial intelligence becomes increasingly common in legal offices and government agencies.
Its use is likely to expand. That makes professional standards and clear guidelines increasingly important.
Courts may eventually need specific policies governing the use of artificial intelligence for legal research, drafting and administrative work. Such policies could require independent verification of AI-generated citations and legal authorities, training on the limitations of generative AI and safeguards for confidential court information.
Technology, however, cannot replace professional judgment. A parent should not be expected to accept a nonexistent court case as legal authority.
A defendant should not bear the consequences of a fabricated statute or citation. A family should not have its legal rights affected by information that was never verified. The responsibility remains with the person who has the authority to make the decision.
The Wheeler matter extends beyond one judge and one court order. It illustrates the growing tension between rapidlyadvancingtechnology and institutions built around careful human judgment.
Courts depend heavily on public confidence. People entering a courtroom must have faith that the law being applied is legitimate, the evidence is being considered and the person making the decision understands the rules governing the case.
Artificial intelligence can assist with the workload; it cannot carry the responsibility. TheWheelerinvestigation could become an important momentinOklahoma'sevolving approach to artificial intelligence in the courtroom.
The lesson is straightforward: Artificial intelligence can make legal research faster and drafting easier but every piece of legal information must still be verified by a human being with the authority and responsibility to get it right. In the courtroom, the final safeguard against artificial intelligence getting the law wrong is still the person behind the bench.