In a case that highlights the perils of overreliance on technology and a lack of due diligence, South Africa-based law firm Surendra Singh and Associates has found itself in hot water for allegedly relying on Artificial Intelligence (AI) to source fictitious legal citations in court proceedings. The controversy has culminated in a sharp rebuke from Judge Elsja-Marie Bezuidenhout of the Pietermaritzburg High Court and a referral to the Legal Practice Council (LPC) for potential investigation.
“Of the nine cases referred to and cited, only two could be found to exist, albeit that the citation of one was incorrect,” Judge Bezuidenhout wrote in her judgment denying an application for leave to appeal. Her remarks came after extensive scrutiny of court documents submitted by the firm on behalf of their client, Godfrey Mvundla, a controversial politician and suspended Mayor of Umvoti.
The firm has been ordered to cover the costs of two hearings held in September 2023, after which Judge Bezuidenhout concluded that the firm’s reliance on AI tools like ChatGPT was likely the source of several non-existent case citations. This, she said, was “irresponsible and downright unprofessional.”
A Chain of Errors
Mvundla’s legal saga began when he contested his suspension as Mayor, securing an interim interdict against the Umvoti Local Municipality. However, Judge Bezuidenhout discharged the interdict and rescinded the order. When Mvundla sought leave to appeal this ruling, the trouble began.
Represented by Ms. S Pillay, counsel for the applicant, the firm’s submissions relied on nine case citations to argue that the judge had erred in law. But when Judge Bezuidenhout’s initial inquiry into one particular citation revealed it did not exist, alarm bells rang. Two law researchers were tasked with verifying the citations in the notice of appeal. Their findings? Seven of the nine cases cited were entirely fictitious, while the remaining two either had incorrect citations or were irrelevant to the matter at hand.
Ms. Pillay, when questioned, attributed the erroneous references to an articled clerk who had drafted the document. The clerk, appearing before the court, claimed to have sourced the cases through a university portal and denied using AI tools. However, her inability to specify the legal journals allegedly consulted raised serious doubts. When tasked with retrieving the cases from a court library, Surendra Singh, the firm’s owner, further muddied the waters by citing technological difficulties and complaining about library fees.
AI Meets Accountability
Judge Bezuidenhout’s investigation into the matter did not stop with the clerk’s explanations. A brief experiment by the court — inputting two of the dubious citations into ChatGPT — illustrated the unreliability of using generative AI tools for legal research. The judge observed that such tools might hallucinate, creating plausible but entirely fictitious legal references.
The implications of this debacle go beyond simple negligence. Judge Bezuidenhout’s ruling emphasized the court’s reliance on counsel to provide accurate citations, describing the firm’s failure as a significant waste of judicial resources. Moreover, the firm’s attempts to deflect blame — with Singh suggesting that the respondent’s counsel was equally culpable for not verifying the citations — did little to curry favor with the court.
Consequences and the LPC’s Role
While the judge refrained from making a definitive ruling on whether AI tools like ChatGPT were used, her comments left little doubt about the gravity of the situation. She directed her judgment to the LPC, urging the regulatory body to review the conduct of the firm and its employees. The LPC, responsible for setting professional standards for legal practitioners, now faces the unenviable task of determining whether disciplinary action is warranted.
Judge Bezuidenhout’s ruling also raised questions about the future legal career of the articled clerk involved, highlighting concerns about her honesty and research methodology. “Presenting fictitious or non-existent cases does not constitute giving an honest account of the law,” she wrote, noting that the clerk’s actions had set off a chain of unfortunate events that culminated in wasted court time and resources.
The Bottom Line
While AI tools hold promise for efficiency, their misuse or overreliance without human oversight can lead to professional embarrassment and legal repercussions. As Judge Bezuidenhout aptly noted, “Had even the most basic checks been conducted, the issue would have been discovered before the document was issued and served.”
For Surendra Singh and Associates, the immediate consequence is financial: the firm must cover the costs of the hearings from its own coffers, even as there are fears the concerned lawyers may be barred from practice.
The full judgment is available here.