Legal practitioners are not going to be replaced by AI. They are going to be replaced by legal practitioners using AI.
The future of legal practice is not to fight the influence of AI, but to use it as a tool to improve productivity and critical analysis. Practitioners now use it to draft, summarise, research, and to prepare for consultations. The temptation to use general-purpose tools such as Claude and ChatGPT is understandable, because they are inexpensive and they produce fluent, confident answers within seconds. However, fluency does not necessarily equate to accuracy in the legal context.
The problem with general-purpose AI
Access to legal databases
Tools such as ChatGPT and Claude are impressive generalists. They are trained on vast quantities of text drawn from the open internet, and they are very good at producing writing that sounds authoritative. The difficulty is that they have no reliable access to South African legal sources. Generalist tools are blocked by sites such as SAFLII, which means they cannot retrieve case law or legislation. Therefore, the AI used can only use secondary sources in its analysis, which jeopardises accuracy. In most cases, when secondary information is not available, generalist AI models will hallucinate.
Risk of hallucination
South Africa has seen its fair share of hallucinated cases. Mavundla v MEC for Co-operative Governance and Traditional Affairs and Northbound Processing (Pty) Ltd v South African Diamond and Precious Metals Regulator provided two examples of the impact of misuse of AI tools. There has also been the recent allegation that an acting judge used AI in a judgement.
Legal privilege
Furthermore, general-purpose AI models may breach legal privilege. Even a pro plan on both ChatGPT and Claude means that your data is being used for retraining. This is undoubtedly a breach of the legal privilege between legal practitioners and their clients. Anthropic (Claude) and ChatGPT have used intellectual property not licensed to them to train their models. A mere prompt requesting the AI to “not use this data for retraining”, does not override its terms and conditions of use, and the AI model will inform you of that fact.
LexCore: AI for South African legal practice
LexCore AI has been designed for South African legal practices that have to incorporate AI into their legal practice, but don't know how. Some of its defining features are:
- Access to the SAFLII database and to South African legislation, which allows it to call on genuine case law and statutory provisions when responding to a query.
- Drafting based on South African precedents, court rules, legislation, and case law.
- Assistants for every matter you deal with.
- Transcription of consultations that can be added to a specific matter's assistant, to be used as context when drafting.
- Integration with all your emails and email management. This includes an email agent — email documents received and request it to be uploaded to an assistant while in court, consultations, etc. Just because you are working does not mean the work has to stop.
Taken together, these are the advantages of a specialised platform: research grounded in real authority, drafting that reflects the legal terminology of the South African jurisdiction, and direct access to the databases that a general tool is blocked from reaching.
Duty of verification
It must be emphasised that no tool, however specialised, relieves the practitioner of the duty to verify. Even a platform trained on South African material has, on occasion, produced errors, and the professional obligation to check every authority before it is placed before a court remains absolute. The proper role of a legal AI is to make the work faster, better informed, and better grounded, and not to replace the independent judgement of the person who signs the opinion.
Conclusion
The risks of using generalist AI models for legal work are clear. A purpose-built South African legal AI platform combines access to information with dedicated features for South African Law. LexCore provides a cost-effective solution for legal practice at an affordable price. Learn more here.
Related blogs
Self-driving cars and Consumer Protection Act liability
What does ChatGPT do with your data?
Deepfakes in South Africa: A practical legal guide
How digital communication is shaping South African law
About the author
Hein Steenberg completed his BCom Law degree, followed by an LLB and LLM in mercantile law, with a specific focus on space law at the North-West University. Hein is currently pursuing his LLD in space law.
Last updated on 31 August 2026.