Artificial Intelligence (AI) is rapidly entering the South African legal sector, promising efficiency in research and contract review. Yet, as local courts are demonstrating, the uncritical integration of AI is not without severe and potentially career-ending risk.
The core problem, known as AI “hallucination,” is where generative AI models invent, distort, or fabricate legal authorities and case citations. While this might seem like a technical glitch, the judicial repercussions in South Africa are immediate and serious, raising urgent questions about professional competence.
South African Courts Show No Tolerance
The High Courts in South Africa are actively scrutinising the use of AI. Judgments from both KwaZulu-Natal and Gauteng serve as non-negotiable warnings:
- The Mavundla Warning: In Mavundla v MEC Department of Co-Operative Government and Traditional Affairs KwaZulu-Natal and Others, the court explicitly stated the lawyer has a duty to act with honesty and competence and to verify the information they use, regardless of the source. The court found it necessary to refer the judgment to the Legal Practice Council (LPC) for attention and further action, sending a clear message that professional negligence in the use of AI will not be overlooked.
- The Northbound Processing Sanction: The Gauteng Division, Johannesburg, in Northbound Processing (Pty) Ltd v The South African Diamond and Precious Metals Regulator and Others, dealt directly with the issue of fabricated citations. Counsel admitted that cases cited in the heads of argument “do not exist” and were seemingly AI “hallucinations.” The presiding Acting Judge directed the Registrar to bring the judgment to the attention of the Gauteng Provincial Office of the LPC for investigation. This is the highest level of judicial scrutiny and potential professional sanction a junior attorney can face.
These cases unequivocally demonstrate that the defence of “the machine made an error” holds no water in a South African court. The legal practitioner, as the final filter, carries the ultimate ethical and professional responsibility for the content submitted.
The Ethical Imperative for Candidate Attorneys
The issue extends beyond fabricated cases to fundamental ethical duties:
- Competence (LPC Code): The duty to act competently means understanding the limitations and risks of the tools you use. Using AI without proper verification is arguably a breach of this duty.
- Candour and Honesty: Presenting fabricated authority, even unintentionally, compromises the practitioner’s duty of honesty to the court.
- Copyright: Legal bodies are also grappling with Copyright and Artificial Intelligence in South Africa, questioning the legality of using copyrighted material to train AI models and the ownership of AI-generated legal work.
The time for cautious observation is over. The challenge is not whether to use AI, but how to use it correctly, ethically, and responsibly to leverage its power while safeguarding your professional integrity.
Secure Your Future: Master the Ethics and Practice of Legal AI
The failure to properly verify AI output can lead to LPC referrals, cost orders (de bonis propriis), and lasting reputational damage.
Our brand new course, How to Use AI Correctly in SA Law, is the systematic guide you need to navigate this new terrain responsibly. We will equip you with the verification strategies, ethical guidelines, and practical workflows demanded by modern South African legal practice, ensuring you move From Candidate to Competent in the digital era.