JUDGE FACES PROBE FOR AI-DRAFTED JUDGMENT WITH FAKE CASES (DOWNLOAD)

The alarming use of artificial intelligence in the legal profession has raised difficult questions about accuracy, professional responsibility and the limits of technological assistance. A recent decision of the High Court of South Africa, Gauteng Division, Johannesburg, has brought one of those questions into sharp focus: what should happen when material apparently generated by artificial intelligence finds its way into a judicial judgment?

The case, F.J.L v T.G.O (Case No. 2025/220239), decided on 31 July 2026, involved a judgment containing a series of erroneous legal authorities, inaccurate quotations and, most significantly, a reference to a case that did not exist. The appellate court considered the possibility that artificial intelligence had been responsible for some of the errors, although it expressly stopped short of making a finding that the Acting Judge had in fact used AI.

The problem with the judgment

The case came before the Full Court on appeal, where counsel drew attention to numerous discrepancies in the judgment of the Acting Judge. Among the problems identified were incorrect citations, quotations that could not be found in the authorities to which they were attributed, passages apparently assembled from different judgments, and references to authorities that did not exist.

One of the most striking examples was a purported authority called Lubbe v Volkswagon SA. The case could not be located, yet it had reportedly been cited several times in the judgment. Other authorities were associated with incorrect case numbers or quotations that did not appear in the judgments cited.

Justice Opperman observed that an AI “hallucination” was the most plausible explanation for some of the fictitious material. Importantly, however, the Court did not find as a fact that the Acting Judge had used artificial intelligence. The judge had not been given an opportunity to respond to the allegation, and the Court therefore treated the matter as one requiring investigation rather than as an established case of judicial AI misuse.

Justice Opperman indicated that the judgment would be forwarded to the Chairperson of the Legal Practice Council for investigation. The Court treated the discrepancies as potentially amounting to professional misconduct, while making clear that its observations were not themselves findings that misconduct had occurred.

The Court formed the view, per the view of Opperman J. that “with the errors or discrepancies in the judgment at face value and without discussing them with the Acting Judge because to do so would be improper. The explanations which I am sure exist are to be addressed to the professional body entrusted with the regulation of the conduct of Acting Judges, not in a private discussion in judicial chambers away from the scrutiny of the litigants and representatives.”

Legal effect of AI written judgment

At paragraph 49 of the decision of the Court, their Lordships held, in the view of Opperman J. that ”I was unable to find any case where AI as a decision-maker was acknowledged or countenanced. I have been unable to identify any reported cases from the UK, Canada, Australia or the US in which a judge expressly delegated legal reasoning or the determination of the outcome of a case to generative AI. Judicial guidance in each jurisdiction strongly suggests that such conduct would be unacceptable because: (a) judicial power is vested in human Judges; (b) Judges have a non-delegable duty to reason and decide cases; (c) litigants are entitled to human adjudication; and (d) judicial accountability would be undermined if reasoning were outsourced to AI.

The Court reasoned, on the effect of the use of AI in determining a matter before a Court that, “Had I been able to conclude that the reasoning in the judgment under consideration was AI-generated and the true decision maker was a machine, I would not have hesitated to conclude that the appellant’s section 34 rights were infringed. To divest the decision-making responsibility to a robot, breaches not only the oath of office, which Acting Judges take,32 but it also denies a party the right to be heard by a Court as intended in terms of section 34.”

The danger of AI hallucinations in the courtroom

Generative AI systems are capable of producing fluent and convincing text while simultaneously generating information that is completely false. The danger is particularly acute in legal research because a fabricated case may look entirely authentic. A citation can contain the correct style, plausible party names, a realistic year and a convincing law-report reference while pointing to no real authority at all. The problem becomes even more serious when the fabricated material is incorporated into a judicial decision.

The Court in the case discusses the treatment of the subject of reliance and use of generative AI in writing judgments and judicial work in common law countries in the West and Asia. A lawyer who cites a fictitious case risks professional sanctions, wasted costs and damage to their credibility. But when a fictitious authority appears in a judgment, the implications extend beyond the individual judge. It raises questions about the integrity of the judicial process and the confidence litigants can place in the courts.

THE JUDGMENT Download

This report originally appeared in THE LAW PLATFORM