South Africa’s Department of Communications and Digital Technologies has published its Draft National Artificial Intelligence Policy for public comment, a comprehensive 86-page framework that for the first time explicitly names the country’s apartheid-era structural inequalities as a primary obstacle to responsible AI adoption.
“Many South Africans remain excluded from digital services due to a lack of connectivity, affordability challenges, or inadequate digital literacy. Moreover, entrenched socio-economic inequality – stemming from apartheid-era structural injustices – continues to affect access to education, employment, and participation in the innovation economy,” the policy draft noted.
The draft, approved by Cabinet on 25 March 2026 and published in the Government Gazette on 10 April 2026, is open for public comment until 10 June 2026.
The policy document, which introduces six strategic pillars including Capacity and Talent Development, Responsible Governance, and Human-Centred Deployment, represents government’s most ambitious attempt yet to harness AI’s estimated $19.9 trillion global economic potential by 2030.
Yet it simultaneously warns that without deliberate intervention, AI risks exacerbating the very inequalities it promises to solve.
Addressing Structural Inequality and the Digital Divide in South Africa’s Tech Sector
South Africa’s AI journey unfolds against a backdrop of persistent structural barriers common to many African nations but uniquely severe due to the country’s apartheid legacy. The draft policy acknowledges that “entrenched socio-economic inequality—stemming from apartheid-era structural injustices—continues to affect access to education, employment, and participation in the innovation economy.”
Across the continent, AI adoption faces three interlocking crises: infrastructure poverty, skills scarcity, and regulatory obsolescence. According to the policy’s own assessment, South Africa suffers from a “persistent digital divide” where underserved and rural communities lack connectivity, affordable devices, and digital literacy.
- The country’s unemployment rate, hovering near 33%, intersects dangerously with automation risks—the policy notes that industries like finance, agriculture, mining, and logistics face potential job displacement without adequate reskilling pathways.
Institutional inertia compounds the problem. South Africa lacks dedicated legislative frameworks for AI-specific risks such as algorithmic bias, accountability in autonomous decision-making, and data protection at scale.
- The policy concedes that “outdated regulatory frameworks would struggle to keep pace with the rapid evolution of AI technologies,” leaving citizens vulnerable while foreign AI providers operate in a legal vacuum.
Across Africa, these challenges are mirrored but magnified by data sovereignty concerns. The policy explicitly warns against “perpetuation of colonial-era data extraction practices,” highlighting how multinational AI firms often extract African data for model training without equitable returns or local accountability.
Also Read: Meta’s AI Glasses Are Recording Private Bedroom Moments – Report
How South Africa’s New AI Policy Framework Drives Ethical Innovation and Growth
The Draft National AI Policy proposes a multi-layered governance architecture unprecedented in Africa.
Central to its approach is the establishment of four new institutions: a National AI Commission to coordinate policy implementation, an AI Ethics Board to enforce fairness and bias mitigation, an AI Regulatory Authority to conduct audits and certifications, and an AI Ombudsperson office allowing citizens to challenge automated decisions.
Crucially, the policy adopts a risk-based approach inspired by the European Union’s AI Act, classifying AI systems by potential harm levels and imposing stricter regulations on high-risk applications in critical infrastructure, healthcare, and finance.
- Unlike purely reactive frameworks, South Africa’s draft creates an AI Insurance Superfund—modelled on the Road Accident Fund—to compensate individuals harmed by AI outcomes where liability proves ambiguous.
On capacity building, the policy mandates AI integration into curricula from primary through tertiary education, including educator training and multilingual instruction. It proposes community-based AI hubs in underserved regions, mobile training centres, and a National AI Skills Development Strategy spanning universities, TVET colleges, and lifelong learning pathways.
- Supercomputing infrastructure, 5G and future 6G networks, low-earth orbit satellites for last-mile connectivity, and regional AI factories are all specified as infrastructure priorities.
The policy’s ‘Futures Triangle’ framework explicitly balances three forces: the ‘Push of the Present’ (technological advancement, economic necessity, social demands), the ‘Pull of the Future’ (economic transformation, social equity, sustainable development, global leadership), and the ‘Weight of the Past’ (digital divide, historical inequities, institutional inertia, outdated regulation). This analytical tool forces policymakers to confront rather than ignore historical baggage.
- Notably, the document enshrines Ubuntu—the African philosophy of interdependence and shared responsibility—as a guiding lens for AI governance, requiring that AI systems “serve the common good, respect diversity, and promote the wellbeing of all.”
It also mandates gender and human rights impact assessments for all public-sector AI deployments, child-centric ethical frameworks limiting data collection from minors, and mandatory explainability for high-stakes systems like credit scoring.
Also Read: Nigerian AI Platform Aims To Solve African Language Barriers With Local Data
Continental Impact: Jobs, Public Services, and Data Sovereignty for Africa
For Africa, South Africa’s draft policy represents a potential blueprint that could influence continental standards through the African Union’s Continental AI Strategy and the African Continental Free Trade Area’s digital trade protocols.
If implemented effectively, the policy positions South Africa not merely as a follower of Western frameworks like the EU AI Act or OECD principles, but as a shaper of “African solutions to African problems” in AI governance.
The jobs impact is substantial but conditional. The policy projects that AI could create new industries and employment categories, but only if the proposed National AI Skills Development Strategy reaches scale.
- South Africa’s youth unemployment crisis—exceeding 45% for those under 35—could see partial relief through the proposed AI accelerators, startup funding mechanisms modelled on Singapore’s mentorship programmes, and tax incentives for AI adoption that prioritises local hiring.
The policy specifically mandates “equitable wages for South African workers in the AI sector who work for foreign employers,” addressing exploitation risks in digital labour markets.
Public service delivery stands to transform most dramatically. The policy identifies healthcare (predictive diagnostics), education (personalised learning platforms), agriculture (precision farming), and public safety as priority sectors for AI deployment. For rural communities where doctor-to-patient ratios are catastrophic, AI-assisted diagnostics could save lives.
For classrooms where teacher shortages are acute, AI tutoring tools operating in indigenous languages could bridge learning gaps. The policy commits to developing real-time translation across all 12 official languages, potentially dissolving communication barriers that have historically marginalised non-English speakers.
Marginalised groups are explicit beneficiaries. The policy mandates “direct involvement of communities affected by disabilities in discussions about AI applications,” investments in assistive AI technologies for children with disabilities, and protection of children from manipulative AI systems like predatory advertising. Women and youth-owned startups receive preferential procurement exemptions from government.
Data sovereignty—a burning issue across Africa—receives unprecedented attention.
- The policy declares that “local personal data provides market intelligence and many other advantages – thus it must be treated as a public good,” requiring compensation for public institutions whose data contributes to AI training.
It activates the National Policy on Data and Cloud to secure cross-border data flows without sovereignty loss, explicitly rejecting “colonial-era data extraction practices.”
For the continent, the stakes extend beyond South Africa’s borders. The policy commits to “federated African cloud/data platforms to enable shared AI resources across Africa for sovereignty and resilience,” and to partnerships with sub-Saharan and continental partners on AI research hubs.
If South Africa successfully implements its proposed AI Ombudsperson, Safety Institute, and Insurance Superfund, these institutions could serve as models for other African nations racing to regulate AI without reinventing governance frameworks from scratch.
The policy’s greatest vulnerability, however, lies in implementation. Its staged approach—Year 1 finalising policy, Year 2 implementing for high-risk cases, Year 3 full deployment—assumes political continuity, sustained funding, and institutional capacity that South Africa’s strained fiscus and uneven governance record may not guarantee.
The document itself admits it is a “work-in-progress” requiring “extensive external consultations with both local and international experts.”
Yet for a continent often relegated to being a consumer of AI technologies developed elsewhere, South Africa’s draft policy makes a radical claim: that African historical realities—inequality, multilingualism, resource constraints, and yes, the weight of the past—can be features, not bugs, in designing AI governance that prioritises human dignity over technological determinism.
Whether this vision survives the 60-day comment period, Cabinet finalisation, and the brutal tests of implementation will determine if South Africa leads or merely legislates.
The draft policy is available for public comment until 10 June 2026. Submissions can be directed to aipolicy@dcdt.gov.za.