- with readers working within the Telecomms industries
- within International Law, Law Department Performance and Consumer Protection topic(s)
Legal Practitioners in South Africa are bound by the Code of Conduct for All Legal Practitioners, Candidate Legal Practitioners and Juristic Entities ("the Code"). At present, the Code does not expressly regulate how legal practitioners can make use of artificial intelligence ("AI") tools. Until the Code is amended, reliance is placed on the three judgments delivered by the South African courts. These judgments have laid the framework regarding the expectations of legal practitioners when using AI tools, particularly in the context of legal research.
Parker v Forsyth N.O. and Others
This matter came before the Johannesburg Regional Court. The plaintiff's legal team sent a list of authorities, without copies of the cases, to the defendant's attorneys, who were unable to find some of the authorities relied upon by the plaintiff. The alleged authorities relied on related to a body corporate's right to sue for defamation.
In court, it emerged that the plaintiff's attorneys had used ChatGPT to conduct legal research but had not verified whether the cases the AI tool generated actually existed. In fact, the cited authorities were fictitious, and the result of AI "hallucinations."
The defendant's counsel argued that this amounted to an attempt to mislead the court and called for a punitive costs order. The court, however, found that the authorities had not been formally submitted to the court as binding precedents, but rather as intended references for the defendant. The court noted that if there had been an attempt to mislead, more serious consequences such as a punitive costs order and a report to the Legal Practice Council ("LPC"), would have followed.
While the court declined to impose such sanctions, it issued a cautionary note: "The efficiency of modern technology still needs to be infused with a dose of good old-fashioned independent reading. Courts expect lawyers to bring a legally independent and questioning mind to bear on, especially, novel legal matters, and certainly not to merely repeat in parrot-fashion, the unverified research of a chatbot".
Mavundla v MEC: Department of Co-Operative Government and Traditional Affairs, KZN and Others
An unsupervised candidate attorney drafted a supplementary notice of application for leave to appeal in a matter before the KwaZulu-Natal Division of the High Court, Pietermaritzburg. In the notice, there were several references to authorities, which allegedly supported the grounds of appeal, based on joinder and service.
A number of the cases relied upon by the candidate attorney, did not exist. The presiding judge gave the applicant's legal team numerous opportunities, by way of appearances in court, to explain why non-existent cases were cited as authorities in the notice. The explanations were unsatisfactory.
ChatGPT was used to conduct the legal research. The court referenced the Code and explained that a court should be able to assume and rely on a practitioner's tacit representation that the authorities cited and relied upon in its papers do actually exist. This goes hand-in-hand with a practitioner's duty to not mislead a court. The use of hallucinated AI-generated legal research could result in the finding that a practitioner intentionally / negligently mislead a court, which would attract professional misconduct charges.
The court explained further that the Code also imposes a duty on a legal practitioner to exercise proper control over their staff and this includes the verification of the accuracy and correctness of any information obtained through the use of AI systems.
It was remarked that "relying on AI technologies when doing legal research is irresponsible and downright unprofessional".
The court ordered the judgment to be brought to the attention of the LPC for further investigation. A costs de bonis propriis (personal costs order) order against the applicant's attorneys in respect of the appearances to verify the existence of the case law was made.
Northbound Processing (Pty) Ltd v South African Diamond and Precious Metals Regulator and Others
Northbound's heads of argument contained hallucinated case law, acting as authority regarding the remedy of mandamus van spolie in a matter that came before the Gauteng Division of the High Court, Johannesburg. The explanation given by Northbound's junior counsel was that the incorrect version of the heads were filed. The legal team for the opposing respondents pointed out that even the correct version of the heads contained two incorrect citations.
The Judge gave Northbound's counsel an opportunity to explain themselves. The junior counsel explained that it appeared that the cited case law was hallucinated by an AI tool. The junior counsel argued that this was not a case similar to Mavundla, because Northbound's Senior Counsel did not rely on the non-existent cases during oral argument, there was no prejudice and the junior counsel accepted full responsibility. The junior counsel disputed that there were attempts to mislead the court.
The court held that it is immaterial that the hallucinated authorities were not referred to during oral argument. What counted is the fact that Northbound's heads of argument contained hallucinated cases, as Judges often rely on heads of argument.
Having quoted the English case of Ayinde v The London Borough of Haringey, the court explained that the following principles apply in South Africa regarding the use of AI tools in the legal profession:
- The risks of using AI to perform legal research are well-known. Those who utilise AI, to conduct legal research, have the professional duty to check the accuracy of the AI-generated research;
- There are serious implications for the administration of justice and public confidence in the justice system, if AI is misused. Practical measures must be taken against the individuals responsible, ensuring that every individual providing legal services understands and complies with their professional and ethical obligations and their duties to the court, if AI is being used;
- The court has a range of powers to ensure that practitioners comply with their duties to the court. The sanctions the court may impose include, but are not limited to, public admonition of the practitioner, adverse costs orders and referral to the regulator behind the legal profession;
- Where a practitioner places false citations before the court due to the use of AI, this is likely a breach of ethical and regulatory requirements, a report to the regulator would be appropriate, with admonishment alone unlikely to be a sufficient sanction.
The court ordered that the judgment to be brought to the attention of the LPC for investigation.
Conclusion
Despite the Code containing no express provisions dealing with AI usage, it is clear that unverified reliance on AI-generated legal research is a breach of a practitioner's ethical obligations and the duties he/she owes to the court. Where hallucinated authorities are before a court, it seems clear that this would constitute an intentional / negligent misleading of the court, depending on the circumstances of the case and the explanation given by the respective legal team. In turn, this could warrant an adverse costs order and/or a report to the LPC for investigation.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
