Skip to Sidebar Skip to Content
Country tracker · Middle East & Africa · 40 countries, one method

AI in South Africa

How far South Africa has come on AI, and how complete its AI rulebook is: two scores, six lenses, every law linked to its official text.
23.9% of people aged 15–64 used a generative AI product in Q2 2026, up from 19.3% in H1 2025; #48 of the 147 economies Microsoft tracks. Its strongest lens is Adoption (#9 of 40) and its weakest Infrastructure (#37). On AI law, a national strategy and voluntary guidance, with no AI law or bill.
Follow South AfricaAI law in South Africa
#33 of 40 on Progress18/100 Rulebook3 verified laws and rulesChecked 21 Oct 2026
In the Record’s memory2 dated events since January 2023, by month and kind. Darker means busier; choose a square to see those events.
See all 2 in the ledger →
Laws and rules2

South Africa through six lenses

How the scores work →
The same six lenses, in the same order, on every country page. Each shows its lead number, the rank among the countries the source covers, and every indicator behind the lens score. 4 of 17 indicators measure South Africa.
01Adoption#9 of 40
How widely are people and firms using AI?
23.9%
share of people aged 15–64 who used generative AI, Q2 2026
H1 2025Q2 2026
Lens score52
03Research#29 of 40
Does it build AI and its science?
0
Below every source's cut-off
Lens score0
05Infrastructure#37 of 40
Does it have the machines?
0
Below every source's cut-off
Lens score0
  • AI computeOct 2026 · Epoch AI GPU Clusters0none recorded
  • Data centres2025 · Cloudscene—source does not cover it
  • Robots installed2024 · International Federation of Robotics (IFR)—below the top 10

AI law in South Africa

How to check a law →

South Africa has no AI law or adopted AI policy; the Cabinet-approved draft National AI Policy was withdrawn in April 2026 after fictitious citations were found.

Strategy and guidanceA national strategy and voluntary guidance, with no AI law or bill.
  1. Revoked or lapsedStrategy · not bindingAdopted 25 Mar 2026

    Draft South Africa National Artificial Intelligence Policy

    Six-pillar draft policy on institutional oversight and sector implementation, gazetted 10 April 2026 for comment; withdrawn by Minister Solly Malatsi on 26 April 2026 after at least six fictitious references were found.

    Official text Independent check Confidence: high
  2. Out for consultationStrategy · not bindingAdopted Aug 2024

    National Artificial Intelligence Policy Framework

    Discussion framework of strategic pillars (talent, digital infrastructure, safety and security, ethics, human-centred approach) intended to lead to a national AI policy and possibly legislation.

    Official text Independent check Confidence: high
  3. In forceLaw · bindingAdopted 26 Nov 2013 · In force from 1 Jul 2020

    Protection of Personal Information Act 4 of 2013 (POPIA), section 71 (automated decision-making)

    Bars decisions with legal or similar effects based solely on automated profiling unless an exception applies (contract, law or code), with a right to make representations.

    Penalty administrative fine up to ZAR 10m; imprisonment up to 10 years for some offences
    Official text Independent check Confidence: high

The Rulebook score, line by line

What each line means →
18 of 100. Each line names the law or body that earns its points.
  1. Binding AI rulesNo AI-specific law, bill or official draft0/35
  2. A national AI strategy, adopted or updated since 2022National Artificial Intelligence Policy Framework5/8
  3. A body named to oversee or enforce AI rulesNone found0/6
  4. An AI safety or evaluation instituteNone found0/6
  5. A right to challenge automated decisions, in forceProtection of Personal Information Act 4 of 2013 (POPIA), section 71…12/12
  6. Rules on AI-generated content: labels or deepfake offences, in forceNone found0/13
  7. International AI governance initiatives endorsed (of 8)1 of 8. Source: AI governance initiatives by country1.5/12
  8. Council of Europe AI Convention: signed 5, ratified 8Not signed0/8

The record

What gets an entry →
Every dated step: laws adopted and coming into force, and each reading of AI use. Newest first; future dates are marked.
  1. 30 Jun 2026AdoptionAI user share reached 23.9%, from 23.1% in Q1 2026Share of people aged 15–64 who used a generative AI product, Q2 2026.Independent estimateSource →
  2. 31 Mar 2026AdoptionAI user share reached 23.1%, from 21.1% in H2 2025Share of people aged 15–64 who used a generative AI product, Q1 2026.Independent estimateSource →
  3. 25 Mar 2026LawsDraft South Africa National Artificial Intelligence Policy: revoked or lapsedSix-pillar draft policy on institutional oversight and sector implementation, gazetted 10 April 2026 for comment; withdrawn by Minister Solly Malatsi on 26 April 2026 after at least six fictitious references were found.Official sourceSource →
  4. 31 Dec 2025AdoptionAI user share reached 21.1%, from 19.3% in H1 2025Share of people aged 15–64 who used a generative AI product, H2 2025.Independent estimateSource →
  5. Aug 2024LawsNational Artificial Intelligence Policy Framework adoptedDiscussion framework of strategic pillars (talent, digital infrastructure, safety and security, ethics, human-centred approach) intended to lead to a national AI policy and possibly legislation.Official sourceSource →
  6. 1 Jul 2020LawsProtection of Personal Information Act 4 of 2013 (POPIA), section 71… came into forceOfficial sourceSource →
  7. 26 Nov 2013LawsProtection of Personal Information Act 4 of 2013 (POPIA), section 71… adoptedBars decisions with legal or similar effects based solely on automated profiling unless an exception applies (contract, law or code), with a right to make representations.Official sourceSource →

How people in South Africa see AI

Public opinion on AI →
Survey results, not scored. They show the mood that new AI rules land in.
64%say AI products have more benefits than drawbacks, 2025+7 points since 2022Ipsos, 2022–25 →55%trust their government to regulate AI responsibly2025Ipsos →64%of employees say they trust AI at work74% use itUniversity of Melbourne and KPMG International →

South Africa beside its peers

All 40 countries →
The two Progress leaders and the countries nearest to South Africa in its region. A dash means no source measures that lens for the country.
MeasureSouth AfricaUnited StatesChinaEgyptNigeriaTurkey
Progressfive lenses, 0–10012#3377#174#210#3710#3816#28
Adoptionlens score525656514846
Talentlens score952—0022
Researchlens score08983000
Investmentlens score010074000
Infrastructurelens score088830014
Rulebook0–100187484143439
AI user shareQ2 2026 · Microsoft23.9%33.0%17.5%15.5%10.4%18.9%
Approach to AI lawStrategy and guidanceState laws, no federal AI statuteBinding rules by use caseStrategy and guidanceAI bill in progressAI bill in progress

More of Middle East & Africa

Every country →
Turkey16Progress39RulebookAI bill in progressOpen tracker →Israel42Progress24RulebookExisting regulators, no AI statuteOpen tracker →Egypt10Progress14RulebookStrategy and guidanceOpen tracker →Saudi Arabia28Progress16RulebookStrategy and guidanceOpen tracker →United Arab Emirates34Progress40RulebookExisting regulators, no AI statuteOpen tracker →Nigeria10Progress34RulebookAI bill in progressOpen tracker →Kenya1Progress34RulebookAI bill in progressOpen tracker →

Check AI-law news yourself

All our data sources →
Every law on these pages links to its official text and to an independent check. These are the registers and trackers we use, with what each is good for. A headline about a new AI law is real when the official register shows it.
Official registers8
Law and policy trackers4
  • OECD.AI Policy Navigator (formerly National AI Policies & Strategies database; GAIIN)OECDNational AI policy initiatives (strategies, governance bodies, funding, regulation) by jurisdictionUse it for Use as primary verification source for 'national AI strategy exists / date' and governance bodies; link each country page to its OECD.AI country profile
  • IAPP Global AI Law and Policy TrackerIAPP (International Association of Privacy Professionals)AI laws, policies, national strategies and responsible authorities per jurisdictionUse it for Cite as a secondary legal cross-check; link out; do not reproduce
  • Council of Europe Framework Convention (CETS 225) treaty chartCouncil of Europe Treaty Office · Live chart (see coe_convention block)Signature and ratification status of the first binding international AI treatyUse it for Primary source for a 'CoE AI Convention: signed / ratified' field per country
  • Digital Policy AlertDigital Policy Alert (St.Gallen Endowment for Prosperity through Trade) · ContinuousEvent-level tracker of digital policy changes incl. AI rules, by jurisdictionUse it for Use as verification source for law-stage changes; model aiadoption.org's country timeline on its event structure
Country indices11
  • Stanford HAI AI Index Report + Global AI Vibrancy ToolStanford Institute for Human-Centered AIAnnual state-of-AI report; Vibrancy Tool ranks national AI ecosystems (R&D, economy, policy, talent, infrastructure, public opinion, etc.)Use it for Reuse individual underlying indicators (with attribution) rather than the composite; link to Vibrancy Tool for each covered country
  • Tortoise Global AI IndexTortoise MediaNational AI capacity: investment, innovation, implementationUse it for Link out and quote rank with attribution only; do not reuse scores
  • Oxford Insights Government AI Readiness Index 2025Oxford InsightsWhether governments can harness AI for the public goodUse it for Cite as a verification source for governance/policy-capacity; reuse rank with attribution for countries lacking other data
  • Microsoft Global AI Diffusion Report (AI user share)Microsoft (AI Economy Institute / On the Issues)AI user share: % of working-age population (15-64) using generative AI products in the periodUse it for Reuse the per-country AI user share as the core 'adoption' metric with attribution and vintage label; flag method breaks
  • IMF AI Preparedness Index (AIPI)International Monetary Fund · June 2024 (released 25 June 2024); no later update foundMacro-structural readiness to adopt AIUse it for Use as background 'structural readiness' component only; label vintage 2024
  • Global Index on Responsible AI (GIRAI)Global Center on AI GovernanceGovernment commitments and implementation evidence for responsible AIUse it for Cite as verification source for responsible-AI governance; adopt its framework-vs-implementation distinction in aiadoption.org's law status pills
  • AI Safety / Security Institutes network (NAAIMES)Member governments; UK AI Security Institute as coordinatorMembership of the international network of AI safety/evaluation institutesUse it for Field 'has national AI evaluation institute (year)' per country; see safety_institutes block
  • Carnegie Endowment (AI governance research, e.g. China AI safety institute work)Carnegie Endowment for International Peace · Ongoing researchQualitative country analysis of AI governanceUse it for Cite for country narrative paragraphs
  • Brookings: How are countries evaluating AI models?Brookings Institution · Article (date not confirmed)Comparative analysis of national AI evaluation institutesUse it for Cite in safety-institute methodology notes
  • Anthropic Economic Index - geographic report (Anthropic AI Usage Index)AnthropicClaude usage per capita relative to working-age population by country (AUI), plus task mixUse it for Use as secondary adoption metric beside Microsoft; triangulate, never sole sourceLink not yet opened by us
  • Epoch AI dataEpoch AI · ContinuousNotable models, training compute, GPU clusters/data centres by countryUse it for Reuse 'notable models by country' and 'AI supercomputers by country' with attributionLink not yet opened by us
Research and safety4
  • CSET (Center for Security and Emerging Technology)Georgetown University · OngoingData-driven research on AI talent, investment, compute, policy (incl. ETO tools such as AGORA AI governance document set)Use it for Cite AGORA for law-text links; reuse open datasets with attributionLink not yet opened by us
  • International AI Safety ReportExpert panel chaired by Yoshua Bengio, UK-backed secretariat · 2026 edition expected/early 2026 (not verified)Scientific synthesis of general-purpose AI capabilities and risksUse it for Cite in methodology for why safety institutes/evaluation capacity are scored; list which countries sit on the panelLink not yet opened by us
  • Our World in Data - Artificial IntelligenceOur World in Data (Global Change Data Lab) · ContinuousCharts republishing AI data (investment, models, patents, public opinion) by countryUse it for Model chart UX on OWID; reuse CC BY datasets with attributionLink not yet opened by us
  • Artificial AnalysisArtificial Analysis · ContinuousModel benchmarks, price and speed; includes model origin by countryUse it for Use for 'frontier models produced by country' cross-check; citeLink not yet opened by us

How the scores work

Two scores, never added together. Download every score and indicator (CSV).

Progress, 0–100

  1. Five lenses, same for everyone. Adoption, Talent, Research, Investment and Infrastructure. Each lens is the average of its indicators; Progress is the average of the lenses.
  2. Only published numbers. Each indicator is a chart on AI Adoption with its source and CSV, or a public dataset linked on the page (Microsoft AI Diffusion, Epoch AI GPU Clusters).
  3. Scaled so outliers do not swamp the rest. Money, models and compute use a log scale on which the leader scores 100; shares and rates use the range of countries the source covers.
  4. Gaps stay gaps. A source that does not cover a country leaves a dash, never a guess. A country below a source's top-15 list scores 0 on that indicator, and the page says so. The coverage flag shows how many of the 17 indicators measure the country.

Rulebook, 0–100

How complete and binding a country's AI rules are. It does not grade whether the rules are good, strict or lenient. Points come only from laws we have verified against an official source.

Binding AI rulesup to 35
  • An AI law with enforceable duties on companies applies now35
  • Enforceable rules on AI systems apply, but only in part of the country26
  • An AI law with enforceable duties is enacted; its duties start later26
  • A national AI law is in force, but it sets no penalties for companies20
  • An AI bill or official draft is before the legislature or out for consultation10
  • No AI-specific law, bill or official draft0
Institutions, safeguards, international65
  • A national AI strategy, adopted or updated since 20228
  • A body named to oversee or enforce AI rules6
  • An AI safety or evaluation institute6
  • A right to challenge automated decisions, in force12
  • Rules on AI-generated content: labels or deepfake offences, in force13
  • International AI governance initiatives endorsed (of 8)12
  • Council of Europe AI Convention: signed 5, ratified 88
Disclosure: AI Adoption uses AI tools in its research and editorial work. No government, company or lobby group has any say in a country's scores or its law register, and the same rules apply to every country. Editorial standards · Data sources · Corrections
Copied with its source. Want the next number like this in your inbox?Get Friday's numbers