SA draft AI policy unveils Superfund and comprehensive oversight framework

While most countries are still debating how to define Artificial Intelligence (AI), South Africa has just come up with a radical challenge.

In a move that could make it a global pioneer, or a cautionary tale, Communications Minister Solly Malatsi's newly gazetted draft national AI policy proposes something unheard of: an AI insurance superfund.

Sample

Modeled after the troubled Road Accident Fund, this “Superfund” is the centerpiece of a policy that refuses to protect it. By suggesting a state-backed financial safety net for victims of algorithmic bias and AI errors, South Africa is acknowledging what Silicon Valley often ignores: AI will cause harm, and someone will have to pay for it.

The 86-page document gazetted on 10 April 2026 signals that the era of “move fast and break things” is over in South Africa.

Policy

This policy doesn't just ask companies to be “ethical”; It creates a formal AI ombudsman and an AI Ethics Board to monitor them.

If an algorithm wrongly denies you a loan or a deepfake ruins your reputation, the state wants a mechanism to ensure you are not left behind in the digital dust.

Malatsi gazetted the draft South African National Cabinet approved Artificial Intelligence policy A 60-day public consultation process formally opens for public comment on Friday.

Malatsi said, “Artificial intelligence presents both significant opportunities and serious risks. When deployed and managed responsibly, AI can help unlock economic growth, improve service delivery, and create new pathways to innovation and inclusion.”

“But if left unchecked, it could also deepen existing inequalities, reinforce exclusion, and expose South Africans to harms arising from technologies developed for very different contexts. The draft national AI policy aims to help South Africa respond to this challenge with a coherent national approach.”

Adoption of AI

Malatsi said the draft policy seeks to support the safe and responsible adoption of AI while protecting rights, promoting inclusive growth and creating conditions for innovation to flourish.

“The policy proposes a framework that balances innovation with opportunity and safeguards with accountability. It is based on South Africa's constitutional values ​​and supports a risk-based approach to governance, with strong monitoring where AI systems pose high risks to people, rights and markets.

Malatsi said, “The draft policy recognizes that a national policy cannot and should not regulate every possible application of AI. Rather, the policy aims to establish core principles that will guide more detailed regional responses over time.”

Counseling

While the current version of the policy builds on feedback received on the South African National Artificial Intelligence Policy Framework and wider intergovernmental engagements, the Department of Communications and Digital Technologies firmly believes that the policy must be shaped through broad consultation if it is to be fit for purpose.

This includes engagement with industry, researchers, civil society, labour, communities and both local and international experts.

new institutions

Part of the policy proposes to create a set of new institutions. These include:

• The National AI Commission (National AI Office) will coordinate reforms on policy, implementation and further policy development with inputs from government, industry and civil society.

• An AI Ethics Board will be established to enforce ethical governance and address considerations related to bias, privacy and fairness.

• The AI ​​Regulatory Authority will monitor compliance, conduct audits and issue certificates.

• The AI ​​Ombudsman Office will allow affected individuals to challenge AI-driven decisions and seek redress.

• The AI ​​Insurance Superfund has been designed on the lines of the Road Accident Fund to compensate individuals or entities harmed by AI-driven outcomes, especially in cases where liability is difficult to determine.

• The National AI Safety Institute will work with similar international bodies to advance the science of AI safety and develop, update, and disseminate guidelines for AI safety as part of a risk and mitigation approach to regulation.

• Integrated AI-powered monitoring center as a central nerve-centre to increase the efficiency levels of all service delivery (and related) touch points in every sector of government and society.

Interference

The draft National AI Policy includes a number of interventions to respond to the safety and security risks posed by AI systems:

Deepfake:

According to the policy, the government should “provide as well as enable clear technical and legal measures against the use of deepfakes and defamatory AI”.

Deepfakes are media – including images, audio, and video – that use artificial intelligence tools to realistically portray a digitally rendered version of a person doing or saying something that never actually happened.

Protecting children from manipulative AI systems:

The draft policy states that AI systems should be prevented from “exploiting the behavioral patterns of children for commercial gain”.

It is clear that children must be protected from AI systems that use predatory advertising and “gamified features that encourage excessive screen time”.

Cyber ​​Security Measures:

According to the draft AL policy, cyber security protocols should be implemented to protect AL systems from unlawful access to data.

With a 60-day period for public comment (ending June 10, 2026), the government is moving at breakneck speed for bureaucracy.

Target? A fully operational AI regulatory framework by the 2027/28 financial year.

According to the draft policy, South Africa's National AI Policy is a strategic blueprint designed to leverage AI technologies for inclusive economic growth, job creation and cost reduction, which will ultimately contribute to Africa's development.

“By prioritizing human-centred AI solutions, the policy aims to improve the quality of life for South Africans and position the country as a leader in AI innovation.”

It is not clear which entity will be tasked with initiating policy interventions once approval is granted.