The lightning deployment of generative AI (GenAI) tools into corporate and legal environments has delivered extraordinary efficiencies. From drafting contracts and summarising board packs to analysing discovery documents, platforms such as ChatGPT and other large language models are now embedded in day-to-day business operations.
Supplied image: Pierre Burger: Partner at Alchemy Law Africa
Yet recent commentary in the United States (US) and the United Kingdom (UK) signals a growing concern: the use of GenAI may place legal professional privilege at risk.
In the US, a decision of a District Court suggested that AI-generated documents may not attract privilege if the communication is not between a lawyer and a client in the traditional sense. Meanwhile, UK commentary has highlighted how the use of GenAI tools could jeopardise claims to privilege if confidential information is disclosed to third-party AI providers or if human legal input is diluted.
For South African businesses and legal practitioners, these developments raise pressing questions. How might GenAI affect legal professional privilege under South African law? And what steps should organisations take to protect themselves?
Legal professional privilege in South Africa: The basics
South African law recognises two primary forms of legal professional privilege:
- legal advice privilege, which protects confidential communications between a client and a legal adviser made for the purpose of obtaining or giving legal advice; and
- litigation privilege, which covers communications between a client or legal adviser and third parties where the dominant purpose is pending or contemplated litigation.
Privilege arises from common law but also underpins constitutional rights by ensuring full and frank disclosure between attorney and client. However, privilege can be waived, either expressly or by conduct inconsistent with its preservation. This is where GenAI becomes relevant.
Melissa Steele 27 Aug 2025 Legal advice privilege and GenAI
Under South African law, legal advice privilege requires a communication between the client and the legal adviser, made in confidence, and for the primary purpose of obtaining or giving legal advice.
If a corporate employee copies legal advice into a public GenAI platform (for example, to simplify it or summarise it), the confidentiality requirement may be compromised. Many GenAI systems are operated by third-party providers and may process, store or learn from user inputs, depending on contractual terms and configurations.
Even if the AI provider states that inputs are not used for training purposes, the transmission of privileged content to an external system may constitute disclosure to a third party, which in South African law generally destroys legal advice privilege unless that third party is an agent necessary for the communication.
The critical question is whether a GenAI tool qualifies as a necessary agent in the privileged relationship or whether it amounts to publishing the advice beyond the protected circle. South African courts have not yet ruled on this issue. However, given the strict requirement of confidentiality, there is a real risk that uploading privileged communications to unsecured or public AI platforms could amount to waiver.
Litigation privilege: A wider shield?
Litigation privilege is broader. It protects communications with third parties, provided litigation is pending or reasonably contemplated, and the dominant purpose of the communication is litigation.
One might argue that using AI to analyse documents for purposes of litigation could fall within this scope. However, the same confidentiality concerns arise. If privileged material is disclosed to an AI provider not acting under the direction and control of legal advisers, and without robust confidentiality safeguards, a court may find that the privilege has been waived.
If GenAI use undermines confidentiality, both legal advice privilege and litigation privilege may be jeopardised. In short, the confidentiality requirement remains a potential point of vulnerability. If GenAI is used in a manner that diffuses confidential legal communications across digital systems outside the control of legal advisers, a court may find that the corporation acted inconsistently with the preservation of privilege.
Where the real danger lies
The risk of losing privilege through GenAI use arises particularly in the uploading of legal advice or draft pleadings to public AI tools; the use of AI platforms without the right safeguards to ensure confidentiality; the processing of privileged material by non-legal employees; the failure to separate privileged and non-privileged content before AI analysis; and reliance on AI-generated summaries of privileged content into non-privileged content.
In each case, the issue is not AI per se, but the potential erosion of confidentiality.
Johnny Davis, Ahmed Dhupli and Simangaliso Sithole 17 Jul 2025 Avoiding the danger
South African businesses should treat GenAI in the same way as they would any other external service provider’s handling of privileged information.
This includes the adoption of clear governance policies around the input of privileged material into public AI platforms; ensuring contractual protections regarding confidentiality, data processing and retention; creating awareness for employees; obtaining legal input into AI adoption strategies; and, where AI is used in litigation support, recording that its use is under the direction of legal advisers for litigation purposes.
Using GenAI cautiously
GenAI is not the enemy of legal professional privilege — but careless use may be. For organisations, the core principles remain unchanged: confidentiality, purpose, and control. What has changed is the ease with which privileged material can be transmitted beyond the protected relationship at the click of a button.
In the AI era, checking your privilege is no longer just a catchphrase. It is a governance imperative.
