South Africa has taken its clearest formal step yet toward regulating artificial intelligence, with government publishing a draft South Africa National Artificial Intelligence Policy for public comment. The draft was approved by Cabinet on 25 March 2026, combined with the special sitting of Cabinet on 1 April 2026, and gazetted on 10 April 2026 by Communications and Digital Technologies Minister Solly Malatsi. Public comments are due by 10 June 2026.
The stronger and more accurate angle here is not that South Africa already has new AI laws in force. The verified position is narrower. Government has opened a draft national AI policy for comment, and that policy explicitly proposes a phased path toward regulatory requirements, guidelines, sector strategies and new institutions over the next three years.
That matters because the draft is not framed as a narrow tech-sector discussion. Cabinet says the policy is meant to help government regulate and adopt AI responsibly while also supporting innovation, job creation and skills development. The policy itself says South Africa risks falling behind, deepening inequality and weakening its ability to protect rights and sovereignty if it does not build a coherent national approach to AI.
What we know so far
The immediate development is the publication of the draft policy in Government Gazette 54477. The gazette notice says interested persons have 60 days to submit written comments, with the final deadline set for 10 June 2026 at 16:00. It also says the draft policy was approved by Cabinet before publication.
Cabinet’s own statement gives the policy a broad national purpose. It says the aim is to ensure that the benefits and risks of AI are distributed fairly across society and generations, and identifies six core pillars: capacity and talent development, AI for inclusive growth and job creation, responsible governance, ethical and inclusive AI, cultural preservation and international integration, and human-centred deployment.
The draft policy makes clear that government is not treating AI as a single-sector issue. It says a national policy cannot and should not address every use of AI directly, and instead should set core principles that guide sectoral approaches. It also says implementation will require a whole-of-government approach, with sector-specific working groups expected to develop roadmaps for areas such as manufacturing, energy, infrastructure, transport and trade.
This is also not the first step in the process. The draft says it builds on the South Africa National Artificial Intelligence Policy Framework published in August 2024, plus 32 submissions received on that framework and later consultations inside government. In other words, the April 2026 publication is a formal policy escalation, not a sudden idea announced without groundwork.
Why it matters
The draft matters because it signals that South Africa is moving from broad AI discussion toward actual governance choices. The policy says current regulatory frameworks may struggle to keep pace with rapid AI development and that legal and institutional reform is needed to support innovation while protecting rights. It specifically situates AI policy within South Africa’s wider legal framework, including the Constitution, POPIA, the Electronic Communications Act, the Electronic Communications and Transactions Act, the Films and Publications Act, the Patents Act, the Cybercrimes Act and PAIA.
The document also shows that government is leaning toward a risk-based model rather than a one-size-fits-all AI law. The staged plan says year one will focus on finalising the policy, identifying key draft regulatory requirements for unacceptable risks, and starting work on national AI policy guidelines. In year two, government plans to publish those guidelines, implement key regulatory requirements for high-risk use cases, and identify draft requirements for medium- and low-risk AI use cases.
That phased approach is important for readers because it means the “new AI laws” story is really about a pipeline of future rules, not one immediate Act that suddenly changes everything on day one. The policy itself says traditional regulatory approaches may not be adequate for AI and argues for a future-proof and technology-neutral model that can adapt as risks differ across sectors.
The draft also signals that government is considering new oversight structures. It proposes an AI Ethics Board, a national AI commission or office, an AI regulatory authority or council, and an AI Ombudsperson office to allow affected individuals to challenge AI-driven decisions and seek redress. The draft further refers to a National AI Safety Institute for high-risk applications and says public-sector and high-risk AI systems should be auditable and contestable.
Key details and figures
Several details stand out in the public record. The draft policy runs to 86 pages. It was gazetted on 10 April 2026 and comments are invited until 10 June 2026. Cabinet approved publication before the notice was issued.
The policy’s timeline is also unusually clear. In 2025/26, government intends to finalise the policy and identify draft regulatory requirements for unacceptable AI risks. In 2026/27, it plans to publish national guidelines, implement key regulatory requirements for high-risk use cases, identify draft requirements for medium- and low-risk systems, develop sectoral AI strategies, and start work on institutional design and funding. In 2027/28, the draft envisages full implementation of remaining policy interventions, with updates as AI technology evolves.
The policy’s content goes beyond general ethics language. It refers to fairness, bias mitigation, privacy, data sovereignty, transparency, explainability and human-rights impact assessments. It says affected individuals should be able to understand and challenge certain AI-driven decisions, especially in high-risk contexts, and it proposes stronger oversight for domains where AI can materially affect people’s rights and opportunities.
Another important point is what the policy does not yet do. It does not publish a final binding licensing regime for all AI tools, and it does not itself amount to a completed standalone AI Act. The draft is still open-ended in places, explicitly describing itself as a work in progress and a point of departure rather than a final settled regulatory model.
What happens next
The immediate next step is the comment process. The gazette notice says comments and written representations must be submitted within 60 days of publication, by 10 June 2026. That means the current phase is public consultation, not final implementation.
After that, the public record suggests the likely sequence will be policy finalisation first, then guidelines and targeted regulatory requirements for higher-risk AI uses, followed by further rules and sector strategies. That is why the most accurate way to describe the current moment is that South Africa has opened the door to new AI rules, rather than already passing a full new AI law.
For readers, the key takeaway is that this is a real and current policy shift. Government is no longer speaking about AI only in broad strategic terms. It has now published a draft national policy that points directly to future regulation, new oversight bodies and a risk-based compliance system, while still leaving room for public input before the framework is finalised.
