There was a time when legal advice came from lawyers, medical advice from doctors, and financial advice from financial advisors. Today, many people consult a chatbot before consulting an expert. Artificial Intelligence (“AI”) has become society’s virtual “Jack of all Trades”, being embraced as a cheaper and faster alternative to professional services. However, cheaper and faster are not always better.
Increasingly, members of the public are turning to AI as a substitute for professional advice. The rise of the “AI lawyer” has cultivated a belief that complex legal disputes and problems can be solved with a carefully worded prompt. While AI has undoubtedly improved access to information, it cannot comprehend, analyse, and advise.
For instance, due to AI’s shortfalls in comprehension, AI cannot distinguish between the applicability of legal principles based on the country where your dispute arose. Following the laws of the wrong country is, however, the least of your concerns, because at least the law might actually exist, albeit elsewhere.
There has been a recent increase in so-called “hallucinated Judgments” generated by AI platforms purporting to provide a legal principle as applied and confirmed by the Courts. Upon further investigation, these Judgments are found to be non-existent. Not even legal professionals are safe from this shortfall.
In Mavundla v MEC, Department of Co-Operative Government and Traditional Affairs, KwaZulu-Natal 2025 (3) SA 534 (KZP), a candidate legal practitioner drafted a Supplementary Notice of Appeal, citing multiple Judgments that either did not exist or did exist but had no bearing on the case before the Court. In Northbound Processing (Pty) Ltd v South African Diamond and Precious Metals Regulator and Others (Leave to Appeal) 2025 JDR 4715 (GJ), the applicant’s heads of argument similarly referred to fictitious Judgments.
Rule 57.1 of the Code of Conduct for all Legal Practitioners, Candidate Legal Practitioners and Juristic Entities imposes a duty on legal practitioners to avoid misleading the Court. In both Mavundla and Northbound, the Judgments were referred to the South African Legal Practice Council for further action.
More recently, there have been allegations of an Acting Judge also referencing hallucinated Judgments in a High Court Judgment.
As seen in the above Judgments, legal representatives who have unfortunately fallen into the manipulative trap of the AI lawyer are held to account by the Judiciary and their regulatory body. The AI lawyer, on the other hand, has no Judge, no regulatory authority, and no Code of Conduct to hold it to account. Similarly, the AI lawyer cannot be sued for professional negligence and has no professional indemnity insurance to protect the layman relying on its (mis)guidance.
The Court in Mavundla correctly stated that:
“In this age of instant gratification, this incident serves as a timely reminder to, at least, the lawyers involved in this matter that when it comes to legal research, the efficiency of modern technology still needs to be infused with a dose of good old-fashioned independent reading. Courts always 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”.
Whilst the statement was made with reference to legal practitioners, it should similarly be a caution to the average person seeking to avoid legal fees, especially in the context of the rise of self-represented litigants, backed by the false confidence of their virtual companion. Whilst AI may be a useful tool in conducting general research that requires simple regeneration of information, it remains a tool. Think twice before consulting the AI lawyer. It may not be worth the “reduced” price after all.