Communications minister Solly Malatsi has gazetted South Africa’s draft nationwide AI coverage for public remark, opening a 60-day window for submissions on an 86-page doc that proposes a sweeping new institutional structure – together with seven new oversight our bodies – to manipulate the event and deployment of AI throughout the economic system.
The draft coverage, revealed within the Authorities Gazette on Thursday, was permitted by cupboard on 25 March. It units out six strategic pillars and dozens of proposed interventions spanning training, digital infrastructure, ethics, knowledge governance and public sector deployment.
However its most hanging characteristic is its bureaucratic ambition. The coverage proposes the creation of:
- A Nationwide AI Fee (or Nationwide AI Workplace) to coordinate coverage and implementation;
- An AI Ethics Board to implement moral governance round bias, privateness and equity;
- An AI Regulatory Authority to conduct audits and concern certifications;
- An AI Ombudsperson Workplace to let residents problem AI-driven selections;
- A Nationwide AI Security Institute;
- An Built-in AI-Powered Monitoring Centre; and, maybe most unusually
- An AI Insurance coverage Superfund modelled on the Highway Accident Fund to compensate people or entities harmed by AI methods the place legal responsibility is troublesome to find out.
As well as, the coverage envisages repositioning communications regulator Icasa for what it calls “an AI-driven regulatory future”, increasing its mandate to supervise moral AI use in telecommunications, ICT and broadcasting.
A brand new Nationwide AI Regulatory Discussion board, coordinated by the division of communications & digital applied sciences, would convey collectively Icasa, the Info Regulator, the Competitors Fee, the South African Reserve Financial institution, the Monetary Sector Conduct Authority, the CSIR and the division of commerce, business & competitors to coordinate oversight.
The size of what’s being proposed raises apparent questions on capability and funding, significantly given authorities’s well-documented struggles to useful resource present establishments. The coverage itself doesn’t connect particular funds figures to any of the proposed our bodies, although it requires funding to be secured in the course of the second yr of a three-year implementation highway map.
Notably, the division of communications has been candid in regards to the doc’s standing. An explanatory notice revealed alongside the coverage describes it as “a piece in progress” and says the federal government’s last strategy “would require in depth exterior consultations with each native and worldwide specialists and curiosity teams”.
‘Present pondering’
The draft “ought to thus be seen as some extent of departure and indication of presidency’s present pondering, relatively than a strict indication of South Africa’s last strategy to the AI coverage panorama”.
This extends to the regulatory strategy itself. The doc presents 4 broad choices – an ethics-first strategy, a versatile iterative mannequin utilizing regulatory sandboxes, an economy-focused technique and alignment with international requirements – and says a mix of all 4, tailor-made to completely different sectors, could be ideally suited. It additionally floats a number of additional choices with out selecting between them: principles-based regulation, a guardrails strategy, a “simply AI” framework targeted on redressing inequality, and even devoted AI laws in sectors the place it’s applicable.
Learn: How AI is reworking the equipment of conflict
The coverage builds on the nationwide AI coverage framework revealed by the communications division in August 2024. It attracts on 32 submissions acquired on that framework and consultations with authorities buildings by the Cupboard cluster course of.
The draft identifies training, healthcare and agriculture because the essential sectors for AI implementation, with public administration as a key lever. It requires AI to be built-in into college curricula from main to tertiary training, for community-based AI training centres to be established in underserved areas and for a labour market transition technique to handle job displacement.
On infrastructure, the coverage requires funding in supercomputing amenities, 5G and future 6G networks, high-capacity fibre and last-mile connectivity by way of low-Earth-orbit satellites. It goes additional, proposing that common web entry be framed as a “socioeconomic proper” and calling for the institution of “regional AI factories” – decentralised compute hubs supposed to advertise native knowledge management and stimulate regional economies.
The coverage additionally proposes that non-private, non-regulated knowledge be handled as a public good and that authorities create incentives for open knowledge initiatives. It requires public broadcasting content material to be made accessible to builders of language fashions, and for AI-powered real-time translation throughout all the nation’s official languages.
Learn: The AI jobs reckoning is right here
The moral dimension of the coverage is grounded within the structure and Invoice of Rights. The doc lists particular constitutional sections that AI should not be used to violate, and frames the African philosophy of “ubuntu” – with its emphasis on interdependence, neighborhood and shared duty – as a guiding lens for AI improvement.
Proposed safeguards
Among the many proposed safeguards are:
- Obligatory human rights and gender influence assessments for AI methods, significantly in high-risk domains;
- Obligatory human-in-the-loop oversight for essential AI selections; and
- Necessities for “adequate explainability” and “adequate transparency” in public sector and high-risk AI methods.
The coverage additionally proposes protections for youngsters towards manipulative AI methods, together with exploitative promoting and gamified options that encourage extreme display time.
Learn: Cupboard approves draft AI coverage for public remark
The division of communications envisions a staged implementation over three years. Within the present monetary yr (2025/2026), the coverage could be finalised, key draft regulatory necessities for “unacceptable dangers” could be recognized and revealed, and work would start on nationwide AI coverage tips.
In yr two (2026/2027), these tips could be revealed, regulatory necessities for high-risk use circumstances could be carried out and sectoral AI methods could be developed. Full implementation is focused for 2027/2028.
The coverage framework would bear a complete evaluation each three years or earlier if triggered by vital technological or legislative shifts.
Written submissions on the draft are due by 10 June 2026. – (c) 2026 NewsCentral Media
Get breaking information from TechCentral on WhatsApp. Enroll right here.
