Country Overview
South Africa does not have a standalone AI act, and its approach to date has combined a sector-by-sector regulatory model with a slow-moving national AI policy process. The Department of Communications and Digital Technologies (DCDT) released an initial AI Policy Framework in October 2024, proposing to lean on existing bodies — the Information Regulator under the Protection of Personal Information Act (POPIA), the Financial Sector Conduct Authority and South African Reserve Bank for financial AI, and the South African Health Products Regulatory Authority (SAHPRA) for medical AI — rather than create a single AI super-regulator, alongside a four-tier risk classification broadly modelled on the EU AI Act.
That process suffered a significant setback in April 2026: after Cabinet published a draft National AI Policy for public comment on April 2, 2026, reporting on April 26 revealed that at least several of the document's roughly 67 references were AI-fabricated, non-existent citations. Minister of Communications and Digital Technologies Solly Malatsi withdrew the draft on April 27, 2026, calling the episode "embarrassing" and pledging consequences for those responsible for drafting and quality assurance. In May 2026, Malatsi appointed an independent expert panel — chaired by Wits University's Professor Benjamin Rosman and including Professor Vukosi Marivate, Professor Alison Gillwald of Research ICT Africa, and others — to authenticate references and guide a redraft, with the revised policy targeted for Cabinet submission by November 2026 and public comment expected around January 2027.
At a Glance
Overall AI policy direction
National AI policy withdrawn and under independent redraft
3
Major AI companies
3
Research institutions
2
Live debates
- Latest development
- July 29, 2026
- Regulatory approach
- Decentralized, sector-based oversight (POPIA/Information Regulator, FSCA, SAHPRA) rather than a single AI-specific regulator, per the withdrawn draft's four-tier risk model
- Major AI companies
- DataProphet, Aerobotics, Lelapa AI
- Major cloud providers
- AWS, Microsoft Azure, Google Cloud
- Leading research institutions
- University of the Witwatersrand (Wits), University of Cape Town (UCT), University of Pretoria
- Key ministries
- Department of Communications and Digital Technologies (DCDT), Information Regulator (South Africa)
- Government involvement
- Active but currently in a reset phase — the DCDT-led national AI policy process was paused mid-2026 pending an independent review
- Industrial strategy
- Emphasis in policy drafts on local datasets and African-language AI processing, alongside applied AI activity in agriculture, manufacturing and language technology
- International partnerships
- Research ICT Africa's Africa Just AI project (funded by IDRC and the FCDO) is a notable example of cross-border AI-policy research collaboration
- Primary policy focus
- Rebuilding the National AI Policy through the independent expert panel process, targeting Cabinet submission by around November 2026
Current Direction
Stable
South Africa's AI governance trajectory was disrupted rather than accelerated in 2026: the government's first attempt at a full National AI Policy was withdrawn after a citation-fabrication scandal, and the policy is now being rebuilt by an independent academic-led panel with a target of reaching Cabinet by around November 2026. In the interim, existing sector regulators (data protection, financial, health) continue to apply general law to AI use cases, and no dedicated AI statute has been introduced in Parliament.
Recent Developments
Following the June 2026 withdrawal of the hallucination-tainted draft National AI Policy, the Department of Communications and Digital Technologies said the redrafted policy -- being rebuilt by an independent expert panel led by Prof Benjamin Rosman -- would take a light-touch approach (29 July 2026): steering sectors to develop their own AI strategies and leveraging existing regulators such as Icasa and the Information Regulator instead of creating seven new institutions. The department targeted Cabinet presentation by November 2026, with January 2027 as a fallback.
2026-07-29SourceTechCentralMalatsi appoints an independent expert panel, chaired by Wits University's Professor Benjamin Rosman and including Professor Vukosi Marivate and Professor Alison Gillwald, to authenticate references and guide a full redraft of the AI policy.
2026-05-12Minister Solly Malatsi withdraws the draft National AI Policy after reporting revealed its reference list contained AI-fabricated, non-existent academic citations.
Major2026-04-27SourceTimesLIVESouth African Cabinet publishes a draft National AI Policy for public comment, built on a four-tier risk model and a decentralized regulatory approach relying on existing sector agencies.
2026-04-02SourceTechCabalDuring its 2025 G20 presidency, South Africa foregrounds AI and digital public infrastructure as headline themes — convening a G20 Task Force on AI, Data Governance and Innovation for Sustainable Development and promoting an 'AI for Africa' agenda — before the rotating presidency passes to the United States for 2026.
2025-11SourceWorld Economic ForumDepartment of Communications and Digital Technologies publishes South Africa's AI Policy Framework, proposing a human-centred, risk-based, sector-led approach as the basis for a future national AI policy.
2024-10-01SourceBowmans
Policy and Regulation
Protection of Personal Information Act (POPIA)
General data-protection law enforced by the Information Regulator; forms the main current legal lever over AI systems that process personal data, including automated decision-making.
National AI Policy Framework
October 2024 discussion document from the DCDT proposing sector-based, risk-tiered AI oversight; a preliminary framework rather than binding law, superseded in status by the since-withdrawn 2026 draft policy.
Draft National AI Policy (2026)
Published for public comment on April 2, 2026, then withdrawn on April 27, 2026, after being found to contain AI-fabricated citations. Being rebuilt by an independent expert panel, with Cabinet submission targeted for around November 2026.
Research Ecosystem
South African AI research is anchored significantly at the University of the Witwatersrand (Wits), whose School of Computer Science and Applied Mathematics and Machine Intelligence and Neural Discovery (MIND) Institute host a notable concentration of AI researchers on the continent, and which hosted the first Deep Learning Indaba in 2017 — a pan-African machine learning conference founded partly in response to the near-total absence of African-institution papers at top ML venues. Research ICT Africa, based in Cape Town and affiliated with the University of Cape Town, is a leading policy-research think tank on AI and digital rights across Africa, led by founding director Professor Alison Gillwald.
Major AI Companies
DataProphet
Industrial AI company providing manufacturing optimization software (PRESCRIBE) aimed at reducing production defects and waste.
Aerobotics
Cape Town-based agritech company using aerial imagery and machine learning to help farmers detect crop pests and disease.
Lelapa AI
Co-founded by Professor Vukosi Marivate, builds resource-efficient language AI models for African languages, addressing gaps in mainstream NLP tooling.
Cloud Infrastructure
- Frontier AI
- No domestic frontier developer
- Compute
- Africa's largest data-centre market (e.g. Teraco); hyperscaler cloud regions and continued AI-infrastructure investment, including a reported Microsoft cloud/AI commitment in 2026
- Chip access
- Import-reliant; no domestic advanced-chip fabrication
AWS
Operates cloud infrastructure serving South African enterprise and startup AI workloads, including a Cape Town region.
Microsoft Azure
Cloud provider with a significant South African enterprise footprint.
Google Cloud
Cloud provider used across South African research and enterprise AI.
Leading Universities
University of the Witwatersrand (Wits)
Hosts the Machine Intelligence and Neural Discovery (MIND) Institute and a School of Computer Science and Applied Mathematics with one of the highest concentrations of AI researchers in Africa; hosted the first Deep Learning Indaba in 2017.
University of Cape Town (UCT)
Affiliated home institution of Research ICT Africa and its AI policy research work.
University of Pretoria
Home to the ABSA-UP Chair of Data Science held by Professor Vukosi Marivate and the African Institute for Data Science and Artificial Intelligence (AfriDSAI).
Think Tanks
Government Institutions
Department of Communications and Digital Technologies (DCDT)
Lead department for South Africa's national AI policy process, headed by Minister Solly Malatsi.
Information Regulator (South Africa)
Primary enforcer of data-privacy law under POPIA; envisaged under the draft AI policy as the lead body for AI-related data-protection issues.
Notable Gaps
Binding national AI policy
South Africa's draft National AI Policy was withdrawn in April 2026 after fabricated citations were discovered; as of mid-2026 there is no adopted national AI policy, only the 2024 preliminary framework and a redraft in progress.
Dedicated AI regulator
South Africa has explicitly chosen not to create a standalone AI regulator, relying instead on existing sector bodies (Information Regulator, FSCA, SAHPRA) — a deliberate design choice rather than an oversight gap, but one whose coordination mechanisms remain unbuilt while the national policy is being redrafted.
Key People
Solly Malatsi
GovernmentMinister of Communications and Digital Technologies
DCDT
Benjamin Rosman
ResearcherProfessor; Chair of the independent expert panel redrafting South Africa's National AI Policy
University of the Witwatersrand
Vukosi Marivate
ResearcherProfessor and ABSA-UP Chair of Data Science; Director, African Institute for Data Science and Artificial Intelligence (AfriDSAI); co-founder, Lelapa AI, Deep Learning Indaba and Masakhane
University of Pretoria
Alison Gillwald
Think tankFounding Director
Research ICT Africa; University of Cape Town
People and Organisations to Follow
Vukosi Marivate
ResearcherLeading South African NLP researcher and AI-for-Africa advocate; co-founder of Lelapa AI, Deep Learning Indaba and Masakhane
Research ICT Africa
Think tankCape Town think tank producing regular AI governance research across Africa
Benjamin Rosman
UniversityWits AI professor chairing the panel rebuilding South Africa's National AI Policy
Current Debates
GovernmentWithdrawal of the draft National AI Policy over AI-fabricated citations and what it means for government AI use
The April 2026 withdrawal of South Africa's draft National AI Policy, after its reference list was found to contain AI-generated fictitious sources, became a widely discussed cautionary case about unverified AI use inside government itself.
▾
Withdrawal of the draft National AI Policy over AI-fabricated citations and what it means for government AI use
The April 2026 withdrawal of South Africa's draft National AI Policy, after its reference list was found to contain AI-generated fictitious sources, became a widely discussed cautionary case about unverified AI use inside government itself.
Minister Malatsi described the episode as a failure of human oversight rather than of the policy's substantive content, and pledged consequences for those responsible for drafting and quality assurance. Commentators, including some South African legal and media outlets, characterised the incident as "embarrassing and damning," and it has fed into the broader debate about how the government can credibly regulate AI risk (including hallucination and verification failures) while its own drafting processes remain vulnerable to the same failure modes.
GovernmentWhether a decentralized, sector-regulator model can substitute for a dedicated AI law
South Africa's policy approach — leaning on the Information Regulator, FSCA, SAHPRA and other existing bodies rather than a single AI authority — is a deliberate design choice that remains contested among policy observers.
▾
Whether a decentralized, sector-regulator model can substitute for a dedicated AI law
South Africa's policy approach — leaning on the Information Regulator, FSCA, SAHPRA and other existing bodies rather than a single AI authority — is a deliberate design choice that remains contested among policy observers.
Proponents argue this avoids duplicating regulatory expertise and lets AI oversight be embedded in bodies that already understand their sectors. Critics, including some of the independent panel members now redrafting the policy, have raised concerns about coordination gaps, unclear escalation paths for cross-sectoral AI harms, and the fact that the coordinating bodies proposed in the withdrawn draft (a National AI Coordination Office, inter-departmental forums, an AI Advisory Council) would lack binding enforcement powers even once established.
Knowledge Hub
Academic papers
Sources
Deliberately sourced beyond government publications. The bars below show each category’s actual share of this page’s citations — an internal editorial guideline rather than a rigorously derived target, so treat it as a rough steer, not a precise benchmark.
- South Africa: Department of Communications and Digital Technologies releases Artificial Intelligence Policy FrameworkThink tanksBowmans2024-10
- Why South Africa is taking a different path to AI regulationNewsTechCabal2026-04-09
- Minister Malatsi withdraws draft AI policyNewsTimesLIVE2026-04-27
- Professor Benjamin Rosman leads South Africa's AI policy panelResearchWits University2026-05
- Malatsi sets up team of experts after AI policy blunderNewsTimesLIVE2026-05-12
- Deep Learning IndabaResearchWikipedia2026
- AI at Wits - The Machine Intelligence and Neural Discovery (MIND) InstituteResearchWits University2026
- Research ICT AfricaThink tanksResearch ICT Africa2026
- DPI and AI are steering South Africa's G20 Presidency agendaThink tanksWorld Economic Forum2025-04
- Radical rethink for South Africa's national AI policyNewsTechCentral2026-07-29