Contact usGet started

Frameworks · South Africa

AI governance frameworks and AI regulation in South Africa

South Africa has no AI-specific law. It governs AI through POPIA, whose section 71 limits decisions based solely on automated profiling, with expectations from the Prudential Authority and the FSCA for financial institutions. The Draft National AI Policy was withdrawn in April 2026 and is being redrafted.

Who regulates AI here

  • Information RegulatorEnforces POPIA, including section 71 on automated decisions, receives security compromise reports and grants prior authorisation for higher-risk processing.
  • Department of Communications and Digital TechnologiesLeads national AI policy, published the National AI Policy Framework in 2024 and is redrafting the National AI Policy.
  • Prudential AuthoritySupervises banks, insurers and market infrastructures, and with the FSCA published the 2025 report on AI in the financial sector and the Joint Standards on IT governance and cybersecurity.
  • Financial Sector Conduct AuthorityRegulates how financial institutions treat customers, including disclosure when AI is used in decisions such as credit assessment and insurance pricing.

South Africa

South Africa regulates automated decisions through its data protection law, POPIA.

LawPOPIA section 71POPIA section 71 and automated decision-makingPOPIA section 71 limits decisions based solely on automated profiling in South Africa. Who it applies to, exceptions, safeguards and what it means for AI.In force since July 2020

All South Africa frameworks

International

Standards and frameworks that apply across borders. Most organisations use one of these as the backbone of their AI governance, whatever their jurisdiction.

StandardISO/IEC 42001ISO/IEC 42001:2023 AI management systemsISO/IEC 42001 is the certifiable international standard for an AI management system. What it requires, how Annex A works and how to get ready to certify.Published December 2023, current editionIn Aicura from ProStandardISO/IEC 42005ISO/IEC 42005:2025 AI system impact assessmentISO/IEC 42005 is international guidance on assessing how an AI system affects people and society, when to assess, what to record and how it links to 42001.Published May 2025, current editionVoluntary frameworkNIST AI RMFNIST AI Risk Management Framework (AI RMF 1.0)The NIST AI RMF is a voluntary framework for managing AI risk through four functions, Govern, Map, Measure and Manage. What it asks and who uses it worldwide.Version 1.0 current, revision under wayIn Aicura from ProVoluntary frameworkNIST Generative AI ProfileNIST AI RMF Generative AI Profile (NIST AI 600-1)NIST AI 600-1 applies the AI Risk Management Framework to generative AI. Its 12 risks, the suggested actions behind them and who uses the profile and why.Final, published July 2024StandardISO/IEC 23894ISO/IEC 23894:2023 Guidance on AI risk managementISO/IEC 23894 adapts the ISO 31000 risk process to AI. What the guidance covers, its AI risk sources and lifecycle mapping, and how it fits with ISO 42001.Published February 2023, current editionStandardISO/IEC 38507ISO/IEC 38507:2022 Governance implications of the use of AI by organizationsISO/IEC 38507 is guidance for boards on governing an organisation's use of AI. What it asks of directors, how it links to ISO/IEC 38500 and who uses it.Published April 2022, current editionVoluntary frameworkOECD AI PrinciplesOECD Recommendation of the Council on Artificial Intelligence (OECD AI Principles)The OECD AI Principles are the intergovernmental standard for trustworthy AI, adopted 2019 and revised 2024. The five principles and how to apply them.Adopted May 2019, revised May 2024Voluntary frameworkHUDERIAHUDERIA: Risk and impact assessment of AI systems for human rights, democracy and the rule of lawHUDERIA is the Council of Europe's method for assessing AI risks to human rights, democracy and the rule of law. Its four elements, scoring and who uses it.Methodology and Model adopted, non-bindingLawCouncil of Europe AI ConventionCouncil of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (CETS No. 225)The Council of Europe AI Convention is the first binding international treaty on AI. Who has signed and ratified, what it asks of parties and when it applies.Open for signature, not yet in forceVoluntary frameworkG7 Hiroshima Code of ConductHiroshima Process International Code of Conduct for Organizations Developing Advanced AI SystemsThe G7 Hiroshima Code of Conduct sets 11 voluntary actions for advanced AI developers, with OECD HAIP reporting. What it asks and who reports under HAIP 2.0.Voluntary, reporting framework v2.0 liveVoluntary frameworkUNESCO AI Ethics RecommendationUNESCO Recommendation on the Ethics of Artificial IntelligenceUNESCO's Recommendation on the Ethics of AI is the global standard its member states adopted in 2021. Its values, principles and Ethical Impact Assessment.Adopted November 2021

All International frameworks

Aicura in South Africa

Frameworks elsewhere

Guides

Questions

Does South Africa have an AI law?

No. The Department of Communications and Digital Technologies published a Draft National AI Policy for comment on 10 April 2026 and withdrew it on 26 April 2026, after its reference list was found to contain sources that do not exist. An independent expert review panel is guiding the redraft. POPIA remains the main binding rule for AI that decides things about people.

Does POPIA regulate AI?

It regulates AI wherever it processes personal information. Section 71 limits decisions with legal consequences or a substantial effect that are based solely on automated profiling, and the rest of POPIA, including notice, security, operator contracts, prior authorisation and offshore transfers, applies to every AI system that touches personal information.

Do financial regulators in South Africa have AI rules?

Not AI-specific ones. The FSCA and the Prudential Authority published a joint report on AI in the financial sector on 24 November 2025, calling for model risk management, board oversight, explainability and disclosure, and said sector-wide guidance on ethical AI is envisaged. Joint Standard 1 of 2023 on IT governance and Joint Standard 2 of 2024 on cybersecurity already apply to AI systems.

Start with the AI you run.

Every framework begins with knowing which AI systems and agents you have. The AI register is free to start.

Sign upContact us