EDITORIAL

 

Call for comments on and withdrawal of the Draft South Africa National Artificial Intelligence Policy

 

 

On 9 April 2026, the Department of Communications and Digital Technologies (DCDT) published the Draft South Africa National Artificial Intelligence (AI) Policy^[1] (hereinafter, the Draft Policy) for public comment, with the final day for submissions being 10 June 2026. Just over two weeks later, on 26 April, the Minister of DCDT withdrew the Draft Policy.^[2] In this editorial I attempt to summarise some aspects of the document, and examine the reasons for withdrawal.

The Draft Policy commences by highlighting how complex a policy agenda for AI is, given the broad range of applications and acceptability of the technology and its rapid development across an overwhelming range of sectors. Accordingly, a general national policy cannot regulate all aspects of AI, and ought not to purport to do so. Its primary purpose is instead to define the overarching principles that anchor and guide differentiated, sector specific approaches. Precisely because AI technologies are broad in scope, cross-cutting and general purpose in character, the implementation of the interventions proposed in the Draft Policy falls within the mandates of a diverse range of government departments and public entities, and effective implementation necessarily requires an integrated, co-ordinated, whole of government approach. AI is positioned as a strategic enabler for South Africa (SA)'s socioeconomic transformation, aligned with the National Development Plan 2030^[3] and national digital transformation goals. The central policy vision is to build an inclusive, ethical, developmental and African centred AI ecosystem that advances economic growth, reduces inequality, and promotes social justice, rather than exacerbating unemployment, bias or digital exclusion. The Draft Policy is based on the South Africa National Artificial Intelligence Policy Framework^[4] of October 2024 and the submissions received on this Framework.

A defining characteristic of the Draft Policy is its developmental framing. Unlike purely market driven AI strategies common in high income countries, SA's policy emphasises:

• AI for poverty reduction, inequality mitigation, and service delivery improvement

• Pro-poor and pro-small and medium enterprise (SME) innovation pathways

• Public interest applications in health, education, agriculture, energy, water, and social protection.

It is stressed that AI adoption must not deepen historical inequities related to race, gender, geography or income. It calls for deliberate interventions to support township, rural and informal sector participation in the AI economy.

Ethics is a major pillar of the Draft Policy. SA commits to a human centred, rights based and constitutional approach to AI governance, grounded in the values of dignity, equality and freedom. Key ethical principles include fairness and non discrimination; transparency and explainability; accountability and human oversight; privacy and data protection; and safety, robustness and security. The policy aligns with international frameworks such as the United Nations Educational, Scientific and Cultural Organization (UNESCO)'s Recommendation on the Ethics of Artificial Intelligence,^[5] but adapts them to SA's constitutional and socioeconomic context. AI adoption in government is framed as a means to improve efficiency and responsiveness, enhance evidence based policy-making, and expand access to services. However, it is stressed that automated decision making in the public sector must include human oversight, transparency, and mechanisms for redress.

The Draft Policy explicitly avoids premature over regulation of AI but calls for adaptive and risk based governance, particularly for high risk use cases (e.g. biometric surveillance, automated decision making affecting rights). Key proposals for a co-ordinated national AI governance framework, rather than fragmented sectoral regulation, include:

• A central AI co-ordinating function within government, led by the DCDT

• Interdepartmental collaboration across economic, social and security portfolios

• Policy coherence with existing laws, including the Protection of Personal Information Act (POPIA), competition law, and labour and employment legislation.

AI is positioned as a catalyst for industrialisation and economic diversification, especially in sectors prioritised by SA's industrial policy, with the draft policy highlighting AI enabled manufacturing and industry; FinTech, GovTech, HealthTech and AgriTech innovation; support for AI start ups and local intellectual property development; and reducing dependence on imported AI solutions. State procurement and public sector demand are identified as tools to stimulate local AI innovation ecosystems, particularly for SA SMEs and research institutions.

AI's disruptive impact on labour markets is acknowledged. Risks of job displacement are noted alongside opportunities for job creation and productivity gains. Key commitments in this context include:

• Large-scale AI, data science, and digital literacy skills development

• Reskilling and upskilling programmes for workers in at-risk sectors

• Curriculum reform from basic education to higher education

• Emphasis on critical thinking, ethics and interdisciplinary skills, not just technical expertise.

The policy aligns AI skills development with youth employment strategies and the digital economy agenda.

Local AI research capacity, particularly in public universities and research councils, is strongly supported, with priority areas including:

• African-relevant AI research (e.g. local languages, health data, climate applications)

• Open science and knowledge sharing

• Multidisciplinary research integrating ethics, law, and social sciences

• International collaboration while safeguarding national interests.

SA is documented as a potential continental AI research hub, contributing to Africa's digital sovereignty.

Recognising that AI depends on high quality data, the Draft Policy prioritises development of national data infrastructure; improved data governance and interoperability; open and shared public sector data, where appropriate; and compliance with POPIA and constitutional privacy protections. Structural challenges such as data fragmentation, quality issues, and unequal access to digital infrastructure, particularly in rural areas, are acknowledged.

Strong emphasis is placed on African and Global South co-operation, with key priorities including:

• Alignment with African Union digital and AI strategies

• South-South collaboration on standards and capacity building

• Strategic engagement with global AI governance processes

• Protecting national interests in global technology markets.

AI is framed not only as a national issue, but as part of Africa's collective digital future.

It is recognised that AI is a rapidly evolving domain, and the Draft Policy commits to ongoing stakeholder consultation; periodic policy review and refinement; evidence based regulation; and flexibility to adapt to technological change. Implementation is envisaged as incremental, collaborative and learning orientated, rather than rigid or "one size fits all".

Having reviewed the Draft Policy and been struck by how polished it appeared, I was therefore not entirely surprised when media reports began to emerge around 25 April indicating that journalists had identified multiple fictitious sources in its reference list, which were almost certainly generated by an AI tool.^[6] Some academic articles cited were completely fictitious and have probably been a hallucination of AI tools used in the research. Ironically, the Draft Policy seems to have been undermined by one of AI's most widely acknowledged limitations. Experts cautioned that the tendency of AI to invent sources stems from its design imperative to always provide an output.^[6]

Acting swiftly on these allegations, the Minister of DCDT announced on 26 April that the Draft Policy had been withdrawn, after initiating internal questions that confirmed the case of hallucinated sources and referencing, compromising the integrity and credibility of the document. He admitted that the Department had not delivered at a "standard that is acceptable for an institution entrusted with the role to lead South Africa's digital policy environment". He further acknowledged that there was no proper verification when including the AI-generated citations. In addition, he remarked that the lapse was unacceptable and illustrated the necessity of robust human oversight over AI. He stated that the matter was being handled with appropriate seriousness and that accountability processes would follow.

It is deeply concerning that a national department, accountable to the SA public, used public funds to publish a national policy in the absence of adequate oversight. Although the Minister's response was prompt, this does not negate the fact that the Department acted irresponsibly, including through the inappropriate use of AI tools. This episode constitutes a stark and cautionary reminder of the inherent limitations and potential dangers associated with the use of AI in the absence of robust human oversight. While AI tools may offer efficiency and support in drafting and research processes, they cannot substitute for professional judgement, accountability and ethical responsibility - particularly in matters of public policy that carry legal, social and moral consequences. The incident highlights how uncritical reliance on AI can undermine institutional credibility, erode public trust, and result in material errors that would probably have been avoided through proper review and governance. It underscores the imperative that AI be deployed as an assistive tool rather than an authoritative source, with clearly defined oversight mechanisms, verification processes and accountability structures firmly in place. Ultimately, this case reinforces the fundamental principle that responsibility for accuracy, integrity and ethical compliance rests with human actors, and that vigilance in this regard is not optional but essential, especially within publicly accountable institutions.

AI statement. AI was utilised to refine language when summarising the Draft Policy.

Ames Dhai

Editor

Ames.Dhai@wits.ac.za

 

References

1. Department of Communications and Digital Technologies, Republic of South Africa. Draft South Africa National Artificial Intelligence (AI) Policy. Government Gazette, Notice 3880 of 2026. https://www.gov.za/sites/default/files/gcis_document/202604/54477gen3880.pdf (accessed 10 April 2026).

2. South African Government News Agency. Minister announces withdrawal of draft AI Policy. https://www.sanews.gov.za/south-africa/minister-announces-withdrawal-draft-ai-policy (accessed 27 April 2026).

3. The Presidency, Republic of South Africa. National Planning Commission. National Development Plan 2030: Our future - make it work. https://www.gov.za/sites/default/files/gcis_document/201409/ndp-2030-our-future-make-it-workr.pdf (accessed 10 April 2026).

4. Department of Communications and Digital Technologies, Republic of South Africa. South Africa National Artificial Intelligence Policy Framework. October 2024. https://www.dcdt.gov.za/sa-national-ai-policy-framework/file/338-sa-national-ai-policy-framework.html (accessed 10 April 2026).

5. United Nations Educational, Scientific and Cultural Organization (UNESCO). Recommendation on the Ethics of Artificial Intelligence. 16 May 2023, last update 26 September 2024. https://www.unesco.org/en/articles/recommendation-ethics-artificial-intelligence (accessed 10 April 2026).

6. Brederode W, Moser K. Govt used fake, made-up research for SA's AI policy. News24, 25 April 2026. https://www.news24.com/business/tech/govts-draft-ai-policy-cites-fictitious-references-experts-believe-are-ai-hallucinations-20260424-1085 (accessed 25 April 2026).