South Africa’s try to determine a nationwide framework for synthetic intelligence (AI) suffered a credibility disaster in April 2026, when the authorities withdrew its draft policy after fabricated educational references have been found in the doc.
The incident uncovered failures in getting ready a framework meant to information the accountable adoption of a expertise already reshaping enterprise and public life.
Five months later, Communications and Digital Technologies Minister Solly Malatsi confirmed that disciplinary proceedings have been ongoing towards one suspended official, following the resignation of one other over the scandal.
“Having such a major stain on an important policy is something that cannot just be allowed to slide,” Malatsi told TechCabal on October 15.
But holding officers accountable is simply half of the problem.
South Africa’s bigger AI governance test is whether or not its current establishments can implement the rules and defend folks from hurt. Strengthening current regulators somewhat than making a devoted AI authority might keep away from pointless forms, however it might place larger accountability on establishments to research AI-related harms, coordinate enforcement, and supply redress. Without clear duties, satisfactory assets and efficient enforcement, a alternative policy dangers changing into a press release of intent somewhat than a framework that adjustments how AI is used.
Flexibility brings an AI enforcement problem
The authorities expects Cabinet to finalise the alternative policy by the finish of March 2027, with public session to observe.
Malatsi acknowledged the delay at the CollectiveXChange 2026 summit, South Africa’s youth employment and digital expertise occasion, in Johannesburg on October 15, describing the course of as a “false start”.
“Yes, it is very late, and that is the consequence of having a false start,” he stated.
The regulatory method can also be evolving. At the X Tech 2026 event series hosted by legislation agency Bowmans on 21 October, Department of Communications and Digital Technologies Deputy Director-General Jeanette Morwane stated the authorities was reviewing laws affected by AI, together with the Films and Publications Act and the Electronic Communications Act.
Morwane indicated that the authorities favoured empowering current regulators over establishing new establishments particularly to supervise AI.
That method might construct on regulators’ sector-specific experience in areas similar to monetary providers, communications and client safety. But distributing oversight throughout establishments creates coordination challenges when AI-related hurt crosses regulatory boundaries.
The alternative policy should make clear which establishment leads investigations, how regulators share info and the way enforcement gaps will likely be addressed. Existing mandates alone is not going to assure a coherent system.
Consumers want greater than moral AI rules
The penalties of weak oversight prolong on to client safety, influencing how people entry important providers and work together with automated enterprise platforms.
At the Bowmans occasion, Heather Irvine, a associate at the agency, famous that banks and insurers routinely use machine studying to evaluate creditworthiness, detect fraud, and value threat, whereas retailers and e-commerce platforms depend on automated personalisation to find out product visibility and pricing.
These functions can enhance effectivity, however in addition they increase questions on accountability when automated choices produce dangerous outcomes.
The dangers are already seen in South Africa’s monetary sector. The South African Banking Risk Information Centre (SABRIC) recorded R2.4 billion ($143.8 million) in losses from digital banking crime in 2025, up from R1.9 billion ($113.8 million) the earlier 12 months. Furthermore, analysis from client credit score company TransUnion highlights how deepfake impersonation and AI-enabled methods are driving increased fraud sophistication throughout Africa.
However, these figures don’t isolate AI-enabled offences, pointing to a key analytical hole: with out devoted information monitoring AI-driven crimes, policymakers will battle to guage the true scale of the menace or decide whether or not current regulators are outfitted to reply.
Can shoppers problem credit score choices influenced by AI? What occurs when personalised pricing disadvantages sure clients? Where can somebody flip when AI-generated content material causes hurt?
Malatsi has argued that moral AI adoption should embody avenues for redress.
“When we create a regime for the ethical use and ethical adoption of AI, where the use is not ethical, there’s got to be an avenue for recourse,” he instructed TechCabal.
The sensible query is how the proposed framework will flip that dedication into accessible cures.
National Consumer Commission performing commissioner Hardin Ratshisusu stated current laws might doubtlessly handle AI-related fraud, misinformation and disinformation. But legal guidelines that will apply to AI-related hurt don’t robotically guarantee regulators can establish violations, examine complaints and ship efficient cures.
The AI capability query can’t be ignored
Morwane highlighted the operational strain going through current our bodies, citing one regulator the place simply two employees members have been tasked with dealing with over 200 circumstances involving baby sexual abuse materials (CSAM).
That instance doesn’t set up that each South African regulator lacks the capability to supervise AI. It does, nevertheless, illustrate why the authorities’s method requires scrutiny.
Effective oversight calls for technical experience, investigative capability and the capacity to evaluate quickly evolving methods. Regulators should additionally decide when current legal guidelines are ample and when amendments or new enforcement mechanisms are wanted.
The alternative policy ought to make clear whether or not regulators will obtain extra funding and technical assist, how their duties will likely be outlined, and the way disputes over jurisdiction will likely be resolved. These will not be minor implementation particulars: they may assist decide whether or not the policy adjustments how AI is developed and deployed.
March is an AI milestone, not the end line
The March 2027 goal presents a milestone after the authentic draft’s withdrawal, however publication alone is not going to restore confidence in AI governance.
The alternative should mix rules for moral adoption and the safety of human rights and dignity with a reputable implementation framework. That means clear institutional duties, timelines, satisfactory assets and procedures for dealing with complaints and enforcing rules.
Public session will give companies, researchers, civil society and shoppers a chance to evaluate whether or not the framework addresses sensible dangers.
South Africa has a motive to keep away from pointless regulatory forms. But flexibility can’t substitute for accountability, assets or enforcement.
The true test will likely be whether or not shoppers can problem dangerous AI choices, companies perceive their compliance obligations, and regulators possess the authorized backing to penalise abuse.
South Africa’s revised AI policy should do greater than restore a flawed drafting course of. It should reveal that when automated methods trigger real-world hurt, accountability extends past moral pledges to establishments with the energy, assets, and authorized mandate to intervene.
