South Africa’s Draft National AI Policy: Building A Framework For Responsible And Inclusive AI Governance

The publication of South Africa’s Draft National Artificial Intelligence (AI) Policy in April 2026 marks an important shift from conceptual discussions on AI toward a structured national governance framework. Rather than proposing immediate, technology‑specific regulation, the draft policy lays down the principles, institutions and implementation pathways that will guide AI development across sectors. It positions AI as a foundational capability that must be aligned with the Constitution, socio‑economic transformation objectives, and South Africa’s long‑term development goals.

For the legal profession, the policy is particularly significant. It anticipates increased reliance on automated decision‑making in both public and private sectors, while emphasising accountability, transparency, and rights protection. In doing so, it signals that law will play a central role in shaping trustworthy AI systems rather than merely responding to them after the fact.

How far is South Africa from Policy Framework to Draft National Policy

South Africa’s AI policy development has been deliberate and phased. Early momentum came from the Presidential Commission on the Fourth Industrial Revolution between 2020 to 2023, which identified AI as a key enabler of inclusive growth. This was followed by the National AI Policy Framework published in August 2024, which articulated high‑level principles and opened the door to extensive stakeholder engagement.

Between 2024 and 2025, public submissions, interdepartmental consultations and international benchmarking informed the transition from that framework to the current draft policy. The publication of the draft in April 2026 therefore reflects not a sudden regulatory impulse, but the maturation of a policy process intended to endure rapid technological change.

The publication of the draft policy on 10 April 2026 has initiated a formal public consultation phase. Written comments are invited for a period of 60 days, with submissions due by 10 June 2026. During this period, government will receive and consider inputs from industry, civil society, academia and the public at large. Once the consultation window closes, the Department of Communications and Digital Technologies is expected to refine the draft in light of those submissions, before progressing it through Cabinet for approval as final policy. Implementation is anticipated to follow on a phased and sector‑specific basis, with further regulatory instruments, institutional arrangements and guidance to be developed over time rather than through a single, immediate legislative intervention.

Core Objectives of the Draft Policy

The draft policy is outcomes‑driven and balances innovation with governance. Its key objectives include (a) leveraging AI for inclusive economic growth and job creation, while recognising the risks of automation; (b) embedding ethical, human‑centred and constitutional values into AI systems across their lifecycle; (c) building national capacity, including skills, infrastructure and public‑sector expertise; (d) actively addressing inequality and the digital divide, ensuring AI does not reinforce historical disadvantage; and (e) positioning South Africa as a credible regional and global participant in AI governance, particularly within Africa.

Crucially, the policy does not treat economic competitiveness, rights protection and social justice as competing interests. Instead, it frames them as mutually reinforcing.

Strategic Pillars and Legal Relevance

The policy is structured around six strategic pillars, all of which carry legal implications.

The Responsible Governance pillar is the most directly relevant to legal practitioners. It introduces a risk‑based approach to AI regulation, anticipates algorithmic audits, and proposes new institutions such as an AI Regulatory Authority and an AI Ombudsperson. These developments will intersect with administrative law, compliance, liability allocation and judicial review of automated decisions.

The Ethical and Inclusive AI pillar grounds AI governance in constitutional rights including equality, dignity, privacy and access to remedies. Concepts such as explainability, human oversight and contestability mirror established administrative‑law principles, translated into the context of algorithmic decision‑making.

From an intellectual property perspective, the Cultural Preservation and International Integration pillar is particularly significant. It recognises the tension between AI innovation and creators’ rights, especially in relation to training data, copyright, performers’ rights and the protection of cultural expression. IP law is thus clearly identified as a key area where AI governance will continue to evolve.

Other pillars covering capacity development, inclusive growth and human‑centred deployment, also raise legal questions concerning professional responsibility, procurement, competition and accountability in public‑sector AI use.

A Phased and Sector‑Specific Implementation Model

A notable strength of the draft policy is its realistic implementation approach. Rather than attempting comprehensive regulation from the outset, it envisages a phased rollout over approximately three years.

Initial efforts focus on governance foundations and high‑risk AI use cases. This is followed by the development of sector‑specific AI strategies, recognising that different sectors from healthcare and education to finance, justice and public administration face distinct risks and opportunities.

Implementation is designed as a whole‑of‑government effort, coordinated by the Department of Communications and Digital Technologies but executed through sectoral roadmaps. Regulatory sandboxes and pilot projects are proposed to enable innovation under supervision, allowing learning before broad regulatory intervention.

Importantly, the aspired new AI‑specific institutions are intended to complement existing regulators rather than duplicate them, fostering coordination without unnecessary regulatory overlap.

International Alignment with Local Context

While grounded in South Africa’s constitutional framework, the draft policy is internationally literate. It reflects alignment with the OECD AI Principles, particularly human‑centred values, transparency and accountability. Its risk‑based regulatory logic also echoes aspects of the EU AI Act, without importing its prescriptive structure wholesale.

Equally important is the policy’s alignment with African Union digital and AI strategies, reinforcing South Africa’s role as both participant and leader in shaping continent‑wide AI governance norms. This approach avoids regulatory imitation, instead favouring interoperability combined with local relevance.

Conclusion: Implications for the Legal Sector

The Draft National AI Policy is more than a precursor to future legislation. It represents a foundational shift toward institutionalised trust in AI governance. For legal practitioners, it signals growing demand for expertise at the intersection of technology, regulation, constitutional rights and intellectual property.

As AI becomes embedded in decision‑making processes, the law will increasingly shape not only accountability after harm occurs, but the design and deployment of AI systems themselves. South Africa’s draft policy makes it clear that AI governance will be human‑centred, constitutionally grounded and development‑oriented, which is a framework that invites sustained engagement from the legal profession.

The SmartAIIP Resource

We have pre-empted this policy through the creation of a unique AI portal and AI readiness scorecard at smartaiip.adams.africa. Conceptually, it has the same staged approach as the RSA policy to AI readiness within organisations. To ensure that you are ready for this, take the following three steps:

1. Complete the AI Readiness Test;
2. Focus on the high risk areas for immediate action and formulate a staged approach over time for the medium and low risk areas; and
3. Contact us, if need be, to find out how our services can assist you achieve.