The Department of Communications and Digital Technologies has confirmed that a re‑draft of South Africa’s national AI policy will not be opened for public comment until January 2027, more than a year after the original document was withdrawn for containing fabricated, likely AI‑generated references. The setback was disclosed to a parliamentary committee on Tuesday, sparking fresh debate about the country’s capacity to steward emerging technologies responsibly.
Minister of Communications Solly Malatsi explained that an internal review failed to spot the erroneous citations, which were first highlighted in a News24 investigation. The minister described the original draft, released in April 2026, as a “starting point” intended to invite stakeholder input. However, the discovery of “massive oversights” and the undisclosed use of AI in compiling the reference list forced the department to pull the paper and launch a remedial process.
To rebuild credibility, Malatsi has appointed a seven‑member independent panel of experts to audit the flawed draft. Acting Deputy Director‑General Jeanette Morwane told the committee that the panel will vet every section, replace dubious sources, and recommend substantive revisions. The revised policy is slated for cabinet submission by November 2026, after which it will be published for public comment in January 2027. Two senior officials have been placed on precautionary suspension while an investigation into the mishap unfolds, Department Director‑General Nonkqubela Jordan‑Dyani calling the episode “highly regrettable”.
The original AI policy aimed to position South Africa as a continental leader in artificial intelligence, balancing ambitious innovation goals with ethical safeguards and economic imperatives. Its withdrawal not only delays those objectives but also raises questions about governance structures, procurement practices and the oversight of AI‑generated content within government ministries.
How the revised AI policy timeline compares to the original schedule
| Milestone | Original Draft (2026) | Revised Draft (2027) |
|---|---|---|
| Public release for comment | April 2026 | January 2027 |
| Parliamentary review | May 2026 | February 2027 |
| Cabinet approval | June 2026 | November 2026 |
| Implementation start date | Q3 2026 | Q2 2027 |
| Stakeholder workshops | July–August 2026 | March–April 2027 |
The table shows a nine‑month delay from the original public consultation period to the new timeline, pushing the policy’s implementation into the second half of 2027. This shift compresses the window for industry and academia to influence the final document, potentially limiting the diversity of input that the government originally sought.
The independent panel, chaired by Prof. Thoko Mahlangu of the University of Pretoria, brings together specialists from academia, the private sector and civil society. Their mandate covers three core areas: ethical governance, economic impact and capacity building. Each expert will submit a detailed report to the department, highlighting where the original draft fell short and proposing concrete amendments.
Key focus areas of the panel’s review
| Domain | Issues flagged in original draft | Proposed enhancements |
|---|---|---|
| Ethical standards | Vague language on bias mitigation; no clear enforcement mechanism | Define measurable bias‑audit protocols; establish an AI Ethics Board |
| Data privacy | Over‑reliance on existing POPIA provisions without AI‑specific clauses | Introduce AI‑tailored data‑handling guidelines and breach reporting |
| Economic incentives | Broad tax‑break promises lacking eligibility criteria | Set transparent grant criteria; create a national AI innovation fund |
| Skills development | No specific targets for up‑skilling the workforce | Mandate university‑industry partnerships; allocate N1 billion to AI training programmes |
The panel’s recommendations aim to substitute the “massive oversight” of the previous version with a robust, accountable framework that can withstand public scrutiny.
Stakeholders have reacted with a mixture of frustration and cautious optimism. The Business Unity South Africa (BUSA) lobby warned that the delay could stall investment pipelines, especially for fintech and agritech firms that rely on clear regulatory guidance. Conversely, the Digital Rights Alliance praised the decision to involve an independent expert body, seeing it as a step toward greater transparency and citizen trust.
Industry players are already positioning themselves for the revised consultation. Local AI start‑ups such as DeepLearnZA and MosaicAI have drafted position papers outlining preferred regulatory approaches, from sandbox environments to co‑funded research initiatives. Meanwhile, multinational corporations operating in South Africa, including Microsoft and Google, have signalled a willingness to collaborate on capacity‑building programmes if the final policy includes clear, predictable incentives.
The parliamentary committee, chaired by MP Thabo Mokgokong, pressed the department for assurances that future drafts will undergo rigorous human verification before release. Morwane affirmed that a new internal audit protocol is being instituted, requiring each citation to be cross‑checked by at least two senior analysts and logged in a transparent ledger. The department also plans to adopt a digital provenance tool that flags AI‑generated text, aiming to prevent a repeat of the current fiasco.
From a broader perspective, South Africa’s AI policy journey reflects a global challenge: governments moving at breakneck speed to harness AI while grappling with the very tools they wish to regulate. The incident underscores the need for meta‑governance—rules that govern the creation and use of policy documents themselves.
As the nation awaits the revised draft, the stakes are high. A well‑crafted AI policy could accelerate the country’s participation in the African AI ecosystem, attract foreign direct investment, and spur job creation in emerging tech hubs like Cape Town’s Silicon Cape and Johannesburg’s Tech Hub. Conversely, continued missteps risk eroding confidence among investors, academia and the public, potentially leaving South Africa lagging behind regional peers such as Kenya and Nigeria, which have already enacted comprehensive AI strategies.
The coming months will test whether the independent panel can deliver a document that not only corrects past errors but also sets a visionary, enforceable standard for AI governance in Africa. If successful, the policy could become a model for other emerging economies navigating the delicate balance between innovation and responsibility. If not, the delay may become another cautionary footnote in the continent’s fast‑evolving digital narrative.