For Communications Minister Solly Malatsi, the withdrawal of the Draft National AI Policy was a sobering necessity, exposing a severe lack of quality assurance in the state’s policymaking machinery.
Image: X / IOLGraphics
THE irony was palpable when South Africa’s Draft National Artificial Intelligence (AI) Policy was abruptly withdrawn on April 26. A document designed to establish the country as an ethical leader in AI governance was found to be riddled with fictitious, AI-hallucinated academic citations.
For Communications Minister Solly Malatsi, the withdrawal of the Draft National AI Policy was a sobering necessity, exposing a severe lack of quality assurance in the state’s policymaking machinery.
The embarrassing debacle, which prompted the appointment of an Independent Expert Review Panel chaired by Professor Benjamin Rosman, serves as a stark reminder: While the state tries to regulate the future using the hallucinations it seeks to police, it loses credibility.
The real AI revolution in South Africa is not happening in committee rooms but is being driven by rapid, decentralised execution on the ground.
This bottom-up adoption is characterised by inclusive usage across various strata of society. A report from the Department of Economics and LEAP at Stellenbosch University shows weekly usage of AI by South African managers and private individuals, with roughly a third leveraging these tools for more than five hours per week.
Critically, adoption is not a luxury of the elite; from a global perspective, usage has reached over 90% among the unemployed, who utilise AI to navigate the labour market. This wallet vote is further evidenced by a 125% explosion in AI subscription payments in 2025, with 40% of credit-card-using South Africans relying on AI for weekly purchase decisions, according to Discovery Spend Trend 2026.
On a corporate level, African business leaders are among the most aggressive global adopters, with 59% of companies preparing to spend over $50 million (R821m) on AI in 2026. This momentum is translating into tangible ROI across key sectors: 71% of finance leaders report that AI is meeting or exceeding expectations, with agentic AI driving a 32% performance improvement.
In industrial and mining sectors, AI is autonomously optimising remote operations at the edge, while the public sector, where South Africa ranks fourth globally in adoption, is using these tools to slash administrative burdens.
Consequently, 82% of managers expect AI to add at least 5% to their productivity over the next three years, moving the technology from pilot projects to an accounting reality.
The newly appointed panel, led by Professor Benjamin Rosman and a team of distinguished experts, faces a clean slate. They must abandon the institutional overhang that doomed the previous draft.
The withdrawn policy was obsessed with the verbs of restriction: regulate, monitor, audit, certify, and enforce. The new policy must instead be anchored on the verbs of growth, diffusion, and reallocation.
To scale AI for all South Africans, the panel must prioritise overcoming physical and regulatory friction, such as planning AI infrastructure around energy resilience due to load shedding (“Energy as Design”) and creating a regulatory framework that enables local inference to satisfy POPIA data residency requirements rather than penalising it.
Ultimately, a national AI policy will be judged by a simple metric: Does it empower South Africans to harness this general-purpose technology, or does it stand in their way?
This requires the panel to adopt an adoption-first approach, which means prioritising infrastructure through cheap bandwidth, last-mile connectivity, and flexible exchange-control rules for software intellectual property.
Furthermore, the state must slash bureaucracy by building national AI measurement capabilities within StatsSA instead of launching heavy-handed monitoring centres, while immediately democratising literacy by embedding AI skills into existing educational programmes.
South Africa’s AI revolution is already happening in boardrooms, on factory floors, and on the smartphones of the unemployed. The true state of AI in 2026 is one of rapid, decentralised execution. The government’s essential job now is to build the runway, stay out of the way, and let the economy take flight.
The balance for the Independent Expert Review Panel is not a zero-sum equation between safety and momentum; rather, safety must be embedded within the growth framework itself.
The pervasive, decentralised adoption of AI (driven by individuals, the unemployed, and aggressive corporate investment) renders traditional, top-down regulatory enforcement obsolete.
* Rowen Pillai (CA (SA), PD (SA)) stands at the intersection of operational excellence and computational intelligence. A Certified AI Expert and MBA graduate, Pillai serves as the co-founder and chief executive of LeanTechnovations.
** The views expressed here do not reflect those of the Sunday Independent, IOL, or Independent Media.