In April 2026 the Minister of Communications and Digital Technologies withdrew South Africa's Draft National Artificial Intelligence Policy, barely three weeks after it had been published for public comment. It emerged that the policy's reference list contained fictitious sources. The episode shows, at the highest level, the risk that has been troubling our courts for some time: artificial intelligence tools can produce material that looks convincing but is simply wrong.
AI is now part of everyday life. Members of the public use it to understand their legal problems, and legal research and drafting platforms increasingly build AI into their services. This article looks at how AI is being used, the risks associated with AI-driven legal platforms, how AI is currently regulated, how the courts and the Legal Practice Council are dealing with its misuse, and why it remains best to seek legal advice.
How Members of the Public Are Using AI
Many AI tools are free or inexpensive and available at any hour. Many people now use them to ask legal questions, to draft letters of demand, to prepare affidavits or to understand documents they have received. Used sensibly, they can help a person understand the basic issues and prepare for a consultation.
There are, however, real risks:
- AI tools can produce answers that are confident but incorrect, including references to cases and legislation that do not exist;
- they may rely on foreign law, or on South African law that has since been amended or overturned;
- they cannot assess the facts of a particular matter, the evidence that is available or the strategy that should be followed;
- information entered into a public AI tool may not remain confidential, and sharing privileged or personal information in this way may have consequences for the user and for others; and
- an AI tool carries no professional responsibility and no insurance if its advice turns out to be wrong.
A person who represents themselves in court is also expected to place accurate information before the court. Relying on an AI-generated document that cites non-existent authority, or that misstates the law, can damage that person's case and credibility.
The Use of AI in Legal Platforms and the Associated Risks
AI is no longer limited to general-purpose AI assistants. Legal research databases, document review tools and drafting platforms now offer AI features that search and summarise judgments, compare documents and produce first drafts. Some newer platforms are marketed specifically as legal AI tools for South African law. These platforms can save time and reduce costs, but they carry risks that every user should understand.
The main risk is what is commonly called a "hallucination". A generative AI tool does not look up the law in the way a law report does. It predicts what a plausible answer looks like. As a result, it can invent case names, citations and quotations that appear entirely genuine. A platform's marketing is no guarantee against this. In the Northbound Processing matter discussed below, the fictitious citations were attributed to an AI legal research tool that claimed to be "exclusively trained on South African legal judgments and legislation".
Other risks associated with AI legal platforms include:
- outdated results, where the platform has not been updated to reflect recent amendments or judgments;
- incorrect summaries, where a real judgment is cited for a proposition it does not support;
- confidentiality and data protection concerns, where client documents or personal information are uploaded to a platform whose data may be stored, used or transferred elsewhere;
- bias in the data on which the tool was trained, which may affect the results it produces; and
- over-reliance, where the convenience of the tool leads users to skip the independent reading and checking that the law requires.
How AI Is Currently Regulated
South Africa does not yet have a statute dealing specifically with artificial intelligence. The Draft National AI Policy, which Cabinet approved in March 2026 and which was published for public comment on 10 April 2026, proposed a phased, risk-based and sector-specific approach to the governance of AI. After the Minister withdrew it, Cabinet's approval of the withdrawal was announced on 5 June 2026 so that the policy could be reworked. A revised draft is expected to be published for comment in early 2027.
In the meantime, the use of AI is governed by existing law. The Protection of Personal Information Act 4 of 2013 applies to personal information processed by AI systems, and section 71 restricts decisions that have legal consequences for a person, or affect them to a substantial degree, and that are based solely on the automated processing of personal information intended to build a profile of that person. Legal practitioners remain bound by the Legal Practice Act 28 of 2014 and the Code of Conduct, which require honesty and integrity and prohibit misleading the court. These duties apply in exactly the same way whether a document was prepared by hand or with the help of AI.
How the Courts Are Dealing with Practitioners
Our courts have made it clear that citing non-existent authority is unacceptable, however it came about.
In Parker v Forsyth NO and Others [2023] ZAGPRD 1 (Johannesburg Regional Court, 29 June 2023) it emerged that authorities cited by the plaintiff's legal representatives did not exist. The court found that the attorneys had relied on an AI tool without checking the results, granted the costs order sought by the defendants and cautioned that "the efficiency of modern technology still needs to be infused with a dose of good old-fashioned independent reading".
In Mavundla v MEC: Department of Co-operative Government and Traditional Affairs, KwaZulu-Natal and Others [2025] ZAKZPHC 2; 2025 (3) SA 534 (KZP), decided on 8 January 2025, the High Court in Pietermaritzburg found that most of the authorities cited in support of an application for leave to appeal, in a supplementary notice of appeal drafted by a candidate legal practitioner, did not exist. The court held that practitioners have a duty to verify the authorities they rely on, ordered the attorneys' firm to pay the costs of the additional court appearances and directed that the judgment be sent to the Legal Practice Council for investigation.
In Northbound Processing (Pty) Ltd v South African Diamond and Precious Metals Regulator and Others [2025] ZAGPJHC 661, decided on 30 June 2025, Smit AJ found that cases cited in the applicant's heads of argument did not exist. Junior counsel confirmed that the citations appeared to be AI hallucinations generated by an AI legal research tool he had used, and senior counsel explained that he had only done a "sense-check" of the heads before they were filed. The court accepted that there was no deliberate attempt to mislead, but stressed at para 92 "the trite duty of legal practitioners not to mislead the court, whether through negligence or intent". At para 95 it warned that "even negligence in this context may have grave repercussions", and it referred the conduct of the applicant's legal practitioners to the Legal Practice Council.
The Role of the Legal Practice Council
The Legal Practice Council regulates attorneys and advocates and may investigate complaints of misconduct. Referral to the Council means that the practitioner's conduct will be investigated and may lead to disciplinary proceedings. Depending on the outcome, the consequences can range from a fine or reprimand to more serious sanctions affecting the practitioner's right to practise.
The message from the courts is consistent. A practitioner who signs a document vouches for the accuracy of the authorities cited in it. Verification is not optional, and reliance on an AI tool is no excuse.
It Is Best to Seek Legal Advice
AI can be a useful starting point, but it is not a substitute for advice from a qualified legal practitioner. An attorney can assess the facts of your matter, apply the current law, identify risks that an AI tool would miss and take professional responsibility for the advice given. If you have used an AI tool to understand your position or to prepare a document, it is best to have a legal practitioner review it before you rely on it, send it or file it at court.
Conclusion
Artificial intelligence is changing the way legal work is done, and its use will continue to grow. It does not, however, change the basic duties owed to the court. Every authority must be real, every quotation accurate and every submission honest. Until South Africa adopts a dedicated AI framework, the courts and the Legal Practice Council will continue to hold practitioners accountable under the existing rules, and members of the public should treat AI-generated legal information with caution.
This article is for general information only and is not legal advice.
