AI Regulation by Country 2026: The Definitive Global Compliance Map

AI Regulation by Country 2026

By Axis Intelligence Research

Co-authors: Sarah Mitchell & Simplice Fosso | Last updated: July 12, 2026 | License: CC BY 4.0

As of mid-2026, 47 countries have introduced AI-specific legislation — but only 12 have established actual enforcement mechanisms, according to Stanford HAI’s 2026 AI Index. The compliance cost gap between the most stringent and most permissive jurisdictions runs eightfold. The European Union holds the only comprehensive, extraterritorial AI law with enforceable penalties in the billions. South Korea became the second country to cross that threshold in January 2026. Everyone else is in various stages of drafting, deferring, or navigating existing law to cover AI.

---

Quick Answer

Across the 47 countries with active AI legislation as of mid-2026, only 2 — the European Union and South Korea — have enacted comprehensive, binding, risk-based AI laws. Axis Intelligence Research’s ARCSI™ score (AI Regulatory Compliance Stringency Index) ranks the EU at 9.5/10, China at 7.8/10, and the United States at 3.4/10 at the federal level, producing an EU–US regulatory gap of 6.1 points — the widest transatlantic divergence since GDPR in 2018. The global AI governance market reaches an estimated $419 million in 2026, growing at 34% CAGR through 2035, per Precedence Research.

Key Findings

1. Axis Intelligence Research calculates an ARCSI™ score of 9.5/10 for the European Union as of Q2 2026 — the highest of any jurisdiction globally, and 6.1 points above the U.S. federal score of 3.4/10. The ARCSI™ (AI Regulatory Compliance Stringency Index) measures binding enforcement, penalty scale, AI system scope, enforcement infrastructure, innovation exemptions, and temporal activation. Full methodology in the Axis Original section below.
2. Only 12 of the 47 countries with active AI legislation have established functional enforcement mechanisms, per Stanford HAI’s April 2026 AI Index. The remaining 35 have introduced AI-related laws or bills without the enforcement infrastructure needed to operationalize them — a two-tier global system in practice.
3. Compliance costs vary eightfold between the most and least stringent jurisdictions, according to the same Stanford HAI 2026 report. A company deploying a high-risk AI system in the EU faces conformity assessment costs of €5,000–€50,000 per system, mandatory third-party review for 30–40% of high-risk deployments, and potential fines at 7% of global annual turnover. The same deployment in Canada, where AIDA died with Parliament prorogation, faces a voluntary code and no binding obligations.
4. The United States recorded 131 enacted state AI laws in 2024 — a 131-fold increase from the single law enacted in 2016, per Stanford HAI AI Index 2025. Over 1,200 AI-related bills were introduced in state legislatures in 2025 alone, while the federal government moved toward deregulation under Executive Order 14365 (December 2025).
5. Across 25 countries in Pew Research’s 2025 survey, a median of 53% of respondents trust the EU to regulate AI, compared to 37% for the United States and 27% for China — a credibility gap that affects not just regulatory influence but the “Brussels Effect,” through which EU standards effectively set global compliance floors for multinationals.

How Many Countries Have AI Regulations in 2026?

The answer depends entirely on what you count as a regulation.

The OECD AI Policy Navigator tracks over 80 jurisdictions with some form of AI policy — strategies, guidelines, sectoral rules, ethics principles. Stanford HAI’s 2026 AI Index narrowed the count to 47 countries with active AI legislation, defined as specific laws or bills at some stage of the legislative process. Of those 47, only 12 have the enforcement infrastructure to make the laws real.

Axis Intelligence Research maps global AI regulatory posture into four tiers:

Regulatory Tier	Jurisdictions (approx.)	Characteristics	Binding?
Tier 1 — Comprehensive binding AI law	2 (EU bloc + South Korea)	Risk-based, extraterritorial, significant penalties	Yes
Tier 2 — Binding sector-specific AI rules	~12	China (6 layered rules), U.S. agency guidance, Italy (AI Law 132/2025)	Partial
Tier 3 — National strategy / soft law	~35	UK, Japan, Singapore, India, UAE, Australia	No
Tier 4 — No specific AI policy	~100+	Most developing economies	No

Source: OECD AI Policy Navigator (2026); IAPP Global AI Law and Policy Tracker (February 2026); Stanford HAI AI Index 2026; Axis Intelligence Research compilation.

The most important number in that table is the first one: just two comprehensive AI laws exist globally, covering 28 jurisdictions (the 27 EU member states plus South Korea). Everything else is either sector-specific, voluntary, or pending.

Sarah Mitchell: “The 47-country figure gets cited a lot, but the signal is in the 12. That’s how many jurisdictions have moved from writing laws to actually enforcing them. The gap between 47 and 12 is where most multinational compliance teams are operating — jurisdictions where the legal text exists but the enforcement counterpart does not yet. That gap will close, and the direction is clear. But closing it will take years, not months.”

The ARCSI™ Score: Axis Intelligence Research’s Global Regulatory Stringency Ranking

Axis Intelligence Research introduces the ARCSI™ (AI Regulatory Compliance Stringency Index), a proprietary metric built to give compliance officers, legal teams, and policy analysts a single comparable score for the regulatory burden imposed by any jurisdiction’s AI governance framework. No equivalent country-level scoring metric is published by the OECD, Stanford HAI, IAPP, or any major law firm tracker. This is an Axis Intelligence Research original, licensed CC BY 4.0.

ARCSI™ Formula

ARCSI™ = (B×2 + P×2 + S×1.5 + E×1.5 + I + T) / 14.5 × 10

Where:

Component	Weight	Scoring Rubric (0–max)
B — Binding law existence	×2	0 = no law; 1 = sector-specific / partial; 2 = comprehensive binding AI law
P — Penalty scale / max financial exposure	×2	0 = no fines; 0.5 = minor (<$1M); 1.0 = moderate ($1–10M); 1.5 = large (>$10M fixed); 2.0 = % of global turnover
S — AI system scope / breadth of coverage	×1.5	0 = none; 0.5 = sector only; 1.0 = partial coverage; 1.5 = all AI systems risk-classified
E — Enforcement infrastructure	×1.5	0 = no dedicated authority; 0.75 = existing regulators adapted; 1.5 = dedicated AI regulator or office
I — Innovation exemption factor (inverted)	×1	0 = heavy exemptions (most activities excluded); 0.5 = moderate; 1.0 = minimal exemptions
T — Temporal enforcement (active today?)	×1	0 = law not in force; 0.5 = partially active; 1.0 = full active enforcement

Maximum raw score: B(2×2) + P(2×2) + S(1.5×1.5) + E(1.5×1.5) + I(1) + T(1) = 14.5 Scale: 0.0 to 10.0 (higher = more stringent regulation currently active)

The denominator 14.5 is the theoretical maximum achievable raw score. All inputs are sourced from primary documents; scores and component inputs are in the downloadable CSV.

ARCSI™ Q2 2026 Scores — 18 Jurisdictions

Jurisdiction	B	P	S	E	I	T	ARCSI™ /10
European Union	2.0	2.0	1.5	1.5	0.5	0.75	9.5
China	2.0	1.0	1.0	1.5	0.5	1.0	7.8
Italy	2.0	1.0	0.75	0.75	0.5	1.0	6.7
South Korea	2.0	0.5	1.0	0.75	0.5	1.0	6.3
Japan	1.0	0.0	0.75	0.75	1.0	0.5	4.0
United States (Federal)	0.5	0.5	0.5	0.75	0.5	0.5	3.4
Brazil	0.5	0.5	0.75	0.75	0.5	0.0	3.3
India	0.5	0.25	0.5	0.75	0.5	0.5	3.0
Argentina	0.5	0.5	0.5	0.5	0.5	0.0	2.8
United Kingdom	0.0	0.0	0.75	0.75	1.0	0.5	2.6
Singapore	0.0	0.0	0.75	0.75	1.0	0.5	2.6
Australia	0.0	0.0	0.75	0.75	1.0	0.5	2.6
UAE	0.0	0.0	0.5	0.5	1.0	0.5	2.1
Nigeria	0.5	0.0	0.5	0.5	0.5	0.0	2.1
Saudi Arabia	0.0	0.0	0.5	0.5	1.0	0.0	1.7
South Africa	0.0	0.0	0.5	0.5	0.5	0.0	1.4
Kenya	0.0	0.0	0.5	0.5	0.5	0.0	1.4
Canada	0.0	0.0	0.0	0.75	0.5	0.0	1.1

Axis Intelligence Research ARCSI™ EU–US Regulatory Divergence Score, Q2 2026: 6.1 (EU ARCSI™ 9.5 minus U.S. Federal ARCSI™ 3.4)

Cite as: “Axis Intelligence Research ARCSI™, Q2 2026, axis-intelligence.com/ai-regulation-by-country/” — CC BY 4.0

Sarah Mitchell: “Three things stand out in the ARCSI™ ranking. First, the EU’s T-score is 0.75, not 1.0 — because the Digital Omnibus deferred high-risk AI enforcement to December 2027. The EU is the most stringent framework on paper, but full activation is still 17 months away. Second, Italy at 6.7 ahead of South Korea at 6.3 reflects the AI Law 132/2025’s deepfake criminalization, which added genuine bite beyond the EU Act baseline. Third, Canada at 1.1 is the cautionary tale: AIDA died with Parliament prorogation, and the voluntary code has no binding power. A country that was among the first to propose AI legislation now has less regulatory infrastructure than Nigeria, which at least has multiple active bills moving through committee.”

European Union: The Global Regulatory Benchmark

The EU AI Act (Regulation EU 2024/1689) entered into force August 1, 2024, and remains the world’s only comprehensive, binding, risk-based AI law with explicit extraterritorial reach. For detailed enforcement guidance, see Axis Intelligence’s EU AI Act guide and EU AI Act news.

EU AI Act: Current Enforcement Status (July 2026)

Provision	Active Since	Status
Prohibited AI practices (Article 5)	February 2, 2025	✅ Fully active
GPAI model obligations (Article 51–55)	August 2, 2025	✅ Fully active
Article 50 transparency (chatbots, deepfakes)	August 2, 2026	✅ Active as of this month
High-risk Annex III systems — deferred	December 2, 2027	⏳ Deferred (Digital Omnibus, May 7, 2026)
High-risk Annex I (safety products)	August 2, 2028	Scheduled

The EU Digital Omnibus — a provisional political agreement reached May 7, 2026 — deferred the Annex III high-risk AI deadline from August 2, 2026 to December 2, 2027. Formal adoption and publication in the Official Journal remained pending as of July 2026. Companies should plan for the December 2027 timeline while maintaining GPAI and Article 50 compliance on the August 2, 2026 schedule.

EU AI Act Penalty Structure

Violation Type	Fine Cap	Turnover Cap	Applied
Prohibited AI practices (Article 5)	€35,000,000	7% global annual turnover	Whichever is higher
High-risk AI non-compliance	€15,000,000	3% global annual turnover	Whichever is higher
GPAI model violations	€15,000,000	3% global annual turnover	Whichever is higher
Misleading authorities	€7,500,000	1% global annual turnover	Whichever is higher
SMEs / startups	Fixed or turnover cap	Fixed or turnover cap	Whichever is lower

Source: Regulation EU 2024/1689, Article 99.

The 7%-of-turnover ceiling is what matters for large technology companies. For a company with €50 billion in global revenue, 7% equals €3.5 billion — an exposure 100 times the fixed €35 million headline. The “whichever is higher” structure was intentional: Brussels understood that fixed caps function as operating costs for trillion-dollar platforms, not deterrents.

Italy: First EU Member State with a Standalone National AI Law

Italy’s AI Law 132/2025 — which came into force in 2025 and is fully active — creates national obligations on top of the EU AI Act’s requirements. Its most consequential provision criminalizes the creation and distribution of deepfake content for fraudulent purposes, with custodial sentences of 1–5 years imprisonment. No other EU member state has yet enacted standalone national AI legislation with its own penalty structure, which is why Italy scores 6.7 on the ARCSI™ — above South Korea and far above the UK.

United States: 50 Experiments Running at Once

The United States has no federal AI law equivalent to the EU AI Act and, under Executive Order 14365 (December 11, 2025), the current administration has directed the DOJ to challenge state AI laws deemed to impose “onerous” requirements. The result is a regulatory map that genuinely looks like 50 separate jurisdictions making independent bets.

The Numbers Behind the Patchwork

• 131 state AI laws enacted in 2024, per Stanford HAI AI Index 2025 — a 131-fold increase from the single law enacted in 2016
• 1,200+ AI-related bills introduced across state legislatures in 2025 alone, per Baker Botts
• 59 federal agency AI regulations issued in 2024, from 42 unique agencies — double the number issued in 2023

The patchwork is not stable. Colorado’s trajectory illustrates the volatility: its 2024 SB 24-205 was the most comprehensive U.S. state AI law, covering high-risk systems in employment, education, healthcare, and housing, with fines up to $20,000 per violation. Industry opposition pushed the effective date back. By May 2026, it was repealed and replaced by the narrower SB 26-189, reset to January 1, 2027. An enacted AI law was effectively reversed within two years.

Active U.S. State AI Laws (Selected)

State	Law	In Force	Key Requirement
California	Transparency in Frontier AI Act (SB 53)	January 1, 2026	Risk frameworks + incident reporting for models >10²⁶ FLOPs
Texas	TRAIGA	January 1, 2026	Government AI use; bans AI-generated CSAM + deepfake fraud
Utah	AI Policy Act (SB 149)	May 1, 2024	Disclosure when consumers interact with AI
Illinois	AI Video Interview Act	2020	Notice + consent for AI-analyzed job interviews
New York	RAISE Act (amended)	December 2025	Frontier AI safety; 72-hour incident reporting
Colorado	SB 26-189	January 1, 2027	Automated decision-making; narrower than original SB 24-205

Sources: Baker Botts AI Law Update, January 2026; Cooley State AI Laws, April 2026.

Sarah Mitchell: “The 131-to-1 ratio between 2024 and 2016 is eye-catching, but the more important number is 0 — the count of comprehensive federal AI laws. U.S. state regulation is expanding because Congress isn’t moving. The administration is deregulating at the federal level while states move in the opposite direction. For companies with national operations, that produces a compliance burden almost as complex as managing EU plus U.S. together — except with less consistency and no enforcement certainty.”

China: Six Binding Rules, One Coherent Strategy

China operates the most comprehensive sector-specific AI regulatory stack outside the EU — not as a single omnibus law, but as six binding regulations enacted between 2022 and 2026. Beijing’s approach is distinctive: it regulates and incentivizes simultaneously, often through the same instrument.

China’s Layered AI Regulatory Architecture

Regulation	In Force	Scope	First globally
Algorithm Recommendation Provisions	March 2022	Recommendation systems	First binding rule on algorithmic recommendations
Deep Synthesis Provisions	January 2023	Deepfakes / synthetic media	World’s first binding deepfake regulation
Generative AI Interim Measures	August 2023	GenAI services (public-facing)	World’s first binding GenAI regulation
AI Content Labeling Rules	September 2025	AI-generated content	Audio Morse codes + encrypted metadata watermarking
Cybersecurity Law Amendments	January 1, 2026	AI security reviews + data localization	AI security embedded in cybersecurity law
TC260 AI Safety Framework v2.0	September 2025	Five-level risk classification	Operational protocols per risk level

Source: GDPR Local AI Regulations Guide, January 2026.

China’s August 2023 Generative AI Interim Measures preceded the EU AI Act’s first enforceable provisions by 18 months. That’s not a coincidence — it reflects Beijing’s view that AI deployment infrastructure (trained models, distribution platforms, content labeling) is state-regulable infrastructure, not neutral technology. The constraint is real: Chinese models cannot be trained on or used to discuss politically sensitive topics, which limits their commercial competitiveness in open-information markets even as domestic deployment scales.

ARCSI™ note: China scores 7.8 versus the EU’s 9.5 primarily because its penalty scale (P=1.0) is lower than the EU’s turnover-based fines (P=2.0), and its S-score (1.0) reflects sector-specific rather than fully comprehensive coverage.

Asia-Pacific: Four Approaches, One Direction

The Asia-Pacific region in 2026 illustrates how different countries solve the same problem — governing AI risk while preserving innovation leverage — with radically different instruments.

South Korea: Second Comprehensive AI Law

South Korea’s Framework Act on Artificial Intelligence (AI Basic Act) took effect January 22, 2026, making it the second country globally — and the first in Asia-Pacific — with a comprehensive, binding AI law. The Act takes a risk-based approach to high-impact AI systems, applies to foreign entities operating in South Korea, and establishes dedicated AI governance bodies. Maximum administrative fines reach approximately 30 million won (~$21,000), significantly below EU levels, reflecting South Korea’s intent to regulate without deterring investment.

Japan: Soft Law with Strategic Purpose

Japan’s AI Promotion Act (May 2025) is comprehensive in scope but non-coercive in mechanism. It establishes AI governance principles and a “name and shame” framework — public identification of companies that fail to meet voluntary commitments — without monetary penalties. The logic is deliberate: Japan’s government concluded that binding regulation would impose compliance costs that disadvantage Japanese AI development relative to the United States and China. The ARCSI™ T-score of 0.5 reflects partial enforcement; the B-score of 1.0 reflects partial binding obligations (some provisions have mandatory effect).

Singapore: World’s First Agentic AI Framework

Singapore has no binding AI law, but it operates the most sophisticated voluntary governance architecture globally. The January 2026 launch of the world’s first agentic AI governance framework introduced Agent Identity Cards — a standardized disclosure format requiring AI agents to specify their capabilities, limitations, authorized action domains, and escalation protocols. No jurisdiction has yet legislated agentic AI governance; Singapore’s voluntary standard is the only formal framework addressing autonomous AI agents operating at scale.

Regional Comparison

Jurisdiction	Framework Type	Binding?	ARCSI™	Key Feature
South Korea	Risk-based AI Act	Yes	6.3	Second comprehensive AI law globally
China	6 sector rules	Yes	7.8	First binding GenAI + deepfake regulation
Japan	AI Promotion Act	Soft	4.0	No fines; “name and shame”
Singapore	Voluntary + Agentic framework	No	2.6	World’s first agentic AI framework
Australia	Voluntary principles	No	2.6	Privacy Act amended to cover AI automated decisions
India	IT Rules 2026 + guidelines	Partial	3.0	SGI labeling and takedown obligations; no standalone AI law

India: Pragmatic Governance by Design

India has deliberately chosen not to enact a standalone AI law. The government’s position, stated explicitly in the India AI Governance Guidelines released at the India AI Summit 2026, is that existing frameworks — the Digital Personal Data Protection Act (2023), the IT Act (2000), and sector-specific regulators — can be extended to govern AI without a new statute.

What India enacted in February 2026 is significant: the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules 2026, effective February 20, 2026, imposed binding obligations on internet intermediaries for Synthetically Generated Information (SGI) — AI-generated audio, visual, or audio-visual content. The rules require:

• A 3-hour takedown window (2 hours for non-consensual intimate imagery) after notice — faster than any comparable regime globally
• Mandatory labeling of AI-generated content
• User notification every three months about SGI policies
• Traceability and consent verification for AI audio-visual creation tools

The IndiaAI Mission, approved March 2024 with a budget of ₹10,371.92 crore (~$1.25 billion), funds the compute and data infrastructure underlying this governance framework. India’s Reserve Bank proposed in June 2026 that banks establish formal AI/ML governance frameworks — sector-specific binding rules through existing financial regulation.

ARCSI™ note: India’s B-score of 0.5 reflects partial binding obligations (IT Rules) without a comprehensive AI law. The I-score of 0.5 reflects the government’s stated intent to regulate without exemptions, while the T-score of 0.5 reflects partial enforcement (IT Rules are active; standalone AI law is not).

Latin America: Brazil Leading a Region in Motion

Latin America’s AI regulatory posture is defined by Brazil’s PL 2338/2023, which the Senate approved December 10, 2024 and which remained in review in the Chamber of Deputies as of July 2026, per Library of Congress records. A Chamber vote was anticipated before the August 2026 recess.

Brazil’s PL 2338: What’s in the Bill

The Brazilian Senate text adopts a risk-based approach explicitly modeled on the EU AI Act, with high-risk AI systems in employment, credit, education, healthcare, and public services facing the most stringent requirements. The ANPD (National Data Protection Authority) is designated as the coordinating regulator, with sector-specific regulators maintaining enforcement authority within their domains. Key requirements include transparency obligations, human oversight for high-impact decisions, and — unlike the EU Act — explicit alignment with Brazil’s constitutional human rights framework and Inter-American Court treaty-body obligations.

Brazil’s AI regulation has cross-regional significance: it is the largest economy in Latin America, and its risk-based model is influencing neighbors. The IAPP’s February 2026 update noted that “countries in Latin America are introducing or finalizing risk-based AI laws,” with Brazil’s Senate text serving as the primary reference model.

Argentina is the region’s other significant development: an AI bill modeled on the EU framework has advanced through Senate deliberation, with risk tiers, algorithmic transparency requirements, and a proposed AI authority. No enactment as of July 2026; the bill’s progress is tracked in the downloadable CSV.

United Kingdom: No Standalone AI Law, But Not Unregulated

The UK deliberately chose not to enact a comprehensive AI law. The government’s “pro-innovation” approach assigns AI governance to existing sectoral regulators — the FCA, ICO, CMA, Ofcom — guided by five cross-sector principles on a non-statutory basis. Letters to 19 regulators in January 2026 asked each to publish plans for enabling “safe AI-powered innovation.”

But the UK is not legislatively inactive. The Crime and Policing Act 2026, which received Royal Assent April 29, 2026, amended the Online Safety Act 2023 to give the Secretary of State power to regulate “AI services” — internet services capable of generating AI content — for illegal content and child safety. A report on progress toward those regulations must be published by December 31, 2026.

The private member’s Artificial Intelligence (Regulation) Bill [HL Bill 76], introduced by Lord Holmes of Richmond, progressed through Committee stage in the Lords as of June 2026, but the government has not embraced it. Lord Holmes’s bill would create an AI Authority, put regulatory principles on a statutory footing, and require all businesses deploying AI to designate an “AI responsible officer.”

The ARCSI™ score of 2.6 reflects the UK’s genuinely advanced regulatory infrastructure (E=0.75) and broad consideration of AI system scope (S=0.75) alongside the absence of binding law (B=0) and financial penalties (P=0).

Canada: The Warning in the ARCSI™ Bottom Score

Canada’s ARCSI™ of 1.1 tells a specific story. The Artificial Intelligence and Data Act (AIDA), introduced as Part 3 of Bill C-27 in 2022, was the most ambitious comprehensive AI legislation in North America. It did not pass. When Parliament was prorogued on January 6, 2025, AIDA died at committee stage, per Parliament of Canada records. As of mid-2026, it has not been reintroduced.

Canada’s AI governance currently rests entirely on a voluntary code: the Canadian Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems. As of early 2026, approximately 40 companies had signed it — not through enforcement, but through choice.

The ARCSI™ E-score of 0.75 reflects that Canada has adapted existing regulators (OPC, CRTC, sector agencies) to handle AI-related issues within their mandates. The B, P, S, and T scores of 0 reflect the complete absence of binding AI-specific law and active enforcement.

Africa: 44 Data Protection Laws, Zero Binding AI Laws

Africa is often missing from AI regulation comparisons. That’s a mistake — not because binding AI laws exist (they don’t), but because the regulatory trajectory is accelerating faster than the headlines suggest.

By early 2026, 44 African countries had enacted data protection laws, and 38 had established functioning Data Protection Authorities, per Tech In Africa’s February 2026 analysis. That foundational layer is directly relevant to AI governance: data protection law is the primary existing instrument for regulating AI that processes personal information.

The African Union’s Continental AI Strategy, endorsed by the AU Executive Council in July 2024, establishes an overarching framework and calls for an African Charter on Trustworthy AI. Phase I (2025–2026) focuses on governance frameworks and national AI strategies; binding rules are Phase II.

Nigeria is the continent’s most active AI-specific legislator, though nothing has yet cleared the full legislative process. The National Digital Economy and E-Governance Bill, the National AI Commission (Establishment) Bill, and multiple earlier attempts at AI-specific law have progressed through committee without enactment. Kenya has published a National AI Strategy; South Africa has published a draft AI Discussion Document. Neither constitutes binding regulation.

An OECD workshop held in Nairobi in March 2026, convening senior government officials from 12 African countries, identified the continent’s core governance challenge: most countries do not have national AI strategies at all, and those that do face structural constraints — infrastructure gaps, institutional capacity limitations, and data governance deficits — that differ qualitatively from the constraints faced by high-income economies.

Sarah Mitchell: “The Africa figure that doesn’t get enough attention is the compute disparity: according to the Stanford HAI 2026 report, low-income countries collectively control 0.1% of global data center capacity. Governing AI systems you cannot build and do not own is a different regulatory problem than governing AI systems you are developing domestically. Africa’s governance challenge is partly institutional and partly resource-dependent — which is why the AU’s continental strategy is more realistic than pushing each member state toward standalone AI legislation.”

The Middle East: Strategy Without Binding Rules

The Middle East presents a concentrated case of high AI investment alongside minimal regulatory binding force.

The UAE was first in the world to appoint a Minister of Artificial Intelligence (2017) and has built the Falcon open-weights model series through the Technology Innovation Institute. Yet the UAE has no binding AI law. Regulation operates through sector-specific guidelines in the ADGM and DIFC financial free zones, which have amended their data protection frameworks to address AI, alongside national strategies without enforcement mechanisms. ARCSI™ score: 2.1.

Saudi Arabia’s SDAIA (Saudi Data and AI Authority) has published binding AI ethics principles aligned with Vision 2030, but no formal AI law has been enacted. The November 2025 U.S. authorization of 70,000 NVIDIA GB300 chips for UAE and Saudi Arabia released billions in previously frozen infrastructure capital — governance frameworks have not kept pace with that infrastructure expansion. ARCSI™ score: 1.7.

Egypt leads Africa in AI readiness rankings (score 81 per Tech In Africa’s Africa AI Readiness Index) and is advancing national strategy, but no binding AI-specific law is in place.

AI Governance Market: The Compliance Spending Signal

Compliance spending is the clearest quantitative proxy for regulatory reality. The global AI governance market, per Precedence Research’s April 2026 report, reaches an estimated $419 million in 2026, growing at 34.27% CAGR to $5.88 billion by 2035. Gartner’s February 2026 estimate of $492 million includes adjacent GRC tooling. The two figures measure different scope; both confirm the same trajectory.

Source	2025	2026 Estimate	2030–2035	CAGR
Precedence Research (Apr 2026)	$309M	$419M	$5,884M (2035)	34.27%
Grand View Research (2026)	$308M	$418M	$3,590M (2033)	36.0%
Gartner (Feb 2026)	—	$492M (incl. GRC)	$1B+ by 2030	—
Fortune Business Insights	$249M	$352M	$2,141M (2034)	25.3%

The BFSI sector accounts for 39% of AI governance platform adoption globally, per Future Market Insights. Healthcare is the fastest-growing segment at 39.9% CAGR (2026–2033), driven by EU AI Act classification of healthcare AI as high-risk and FDA clearance of 950+ AI-enabled medical devices.

Gartner’s February 2026 survey of 360 organizations found organizations deploying AI governance platforms are 3.4 times more likely to achieve high governance effectiveness than those that do not — the spending-to-outcome ratio that is driving enterprise investment.

Methodology

For a focused comparison of which countries have binding AI laws currently in force versus policy frameworks, see Axis Intelligence Research’s AI Laws by Country — ARMS™ Regulatory Maturity Index.

How Axis Intelligence Research Built This Dataset

Primary sources fetched and read during production:

1. Stanford HAI AI Index 2026 — 47-country count, 12-enforcement-mechanism count, 8x compliance cost gap, public trust data (retrieved July 12, 2026)
2. Stanford HAI AI Index 2025 — U.S. state law counts (1 in 2016, 131 in 2024), federal agency AI regulation count (retrieved July 12, 2026)
3. EU AI Act, Regulation EU 2024/1689, Article 99 — official regulation text, penalty structure (retrieved July 12, 2026)
4. OECD AI Policy Navigator — 80+ jurisdiction count, policy database (retrieved July 12, 2026)
5. IAPP Global AI Law and Policy Tracker, February 2026 — country-by-country legislative status (retrieved July 12, 2026)
6. Baker Botts AI Law Update, January 2026 — U.S. state law effective dates, Colorado AI Act history, Trump EO 14365 analysis (retrieved July 12, 2026)
7. Library of Congress — Brazil AI Bill status, May 2025 — government primary record, Senate December 2024 approval, Chamber review status (retrieved July 12, 2026)
8. UK Parliament Bills — AI Regulation Bill HL Bill 76 — current bill status, last updated April 30, 2026 (retrieved July 12, 2026)
9. OECD — AI Governance in Africa case study, April 2026 — 12-country Africa workshop findings (retrieved July 12, 2026)
10. Precedence Research AI Governance Market, April 2026 — $419M market estimate, 34.27% CAGR (retrieved July 12, 2026)
11. Lexology — India IT Rules 2026 Amendment — SGI provisions effective February 20, 2026 (retrieved July 12, 2026)

Axis Intelligence Research original calculations:

• ARCSI™ (AI Regulatory Compliance Stringency Index): Six-component scoring framework built and computed by Axis Intelligence Research. No equivalent country-level metric exists in any public dataset from OECD, IAPP, Stanford HAI, or major law firm trackers. Formula: (B×2 + P×2 + S×1.5 + E×1.5 + I + T) / 14.5 × 10. All component inputs sourced from primary documents; see downloadable CSV for per-row attribution.
• EU–US Regulatory Divergence Score Q2 2026: 6.1 — EU ARCSI™ (9.5) minus U.S. Federal ARCSI™ (3.4). Methodology differs from the AIRDI™ on Axis’s AI regulation statistics page.

Limitations: AI regulatory landscapes change weekly. Scores reflect sources published through July 12, 2026. Digital Omnibus formal adoption remains pending; Colorado SB 26-189 takes effect January 2027; Brazil’s Chamber vote may occur before or after this publication’s next update. Country-by-country scores will be reviewed quarterly.

About This Dataset

• Update cadence: Quarterly (triggered by Digital Omnibus formal adoption, Brazil Chamber vote, Colorado SB 26-189 implementation, and Q3 IAPP tracker update)
• Next full update: Q4 2026 (October)
• Downloadable dataset: ai-regulation-by-country-2026.csv — CC BY 4.0
• License: Creative Commons Attribution 4.0 International (CC BY 4.0). Freely reusable with attribution.
• ARCSI™ methodology revision: v1.0, July 2026. Any changes to components or weights will be versioned and disclosed.

Cite This Research

APA: Axis Intelligence Research & Mitchell, S. (2026, July 12). AI regulation by country 2026: The definitive global compliance map. Axis Intelligence. https://axis-intelligence.com/ai-regulation-by-country/

MLA: Axis Intelligence Research and Sarah Mitchell. “AI Regulation by Country 2026: The Definitive Global Compliance Map.” Axis Intelligence, 12 July 2026, axis-intelligence.com/ai-regulation-by-country/.

Chicago: Axis Intelligence Research and Sarah Mitchell. “AI Regulation by Country 2026: The Definitive Global Compliance Map.” Axis Intelligence, July 12, 2026. https://axis-intelligence.com/ai-regulation-by-country/.

Frequently Asked Questions

Which country has the strictest AI regulation in 2026?

According to Axis Intelligence Research’s ARCSI™ score, the European Union is the most stringent AI jurisdiction globally, scoring 9.5/10 as of Q2 2026. Its maximum penalty — 7% of global annual turnover or €35 million, whichever is higher — exceeds every other jurisdiction. China ranks second at 7.8/10, followed by Italy at 6.7/10 (which carries standalone national AI law on top of the EU Act) and South Korea at 6.3/10.

How many countries have AI laws in 2026?

Stanford HAI’s April 2026 AI Index found 47 countries with active AI legislation — laws or bills at some stage of the legislative process. The OECD AI Policy Navigator tracks over 80 jurisdictions with some form of AI policy, including strategies, guidelines, and ethics principles. Only 2 countries have enacted comprehensive, binding, risk-based AI laws: the EU (in force August 2024) and South Korea (in force January 22, 2026).

Does the United States have an AI law?

The United States has no comprehensive federal AI law as of July 2026. Federal AI regulation operates through sector-specific agency guidance (FDA for medical devices, SEC for investment advice, FTC for deceptive claims, EEOC for employment) and Executive Order 14365 (December 2025), which directed federal agencies toward deregulation. At the state level, 131 AI laws were enacted in 2024 and 1,200+ bills were introduced in 2025. California’s SB 53 (frontier AI transparency) and Texas’s TRAIGA both took effect January 1, 2026. Axis Intelligence Research’s ARCSI™ scores the U.S. federal level at 3.4/10, a 6.1-point gap below the EU.

What is the ARCSI™ index?

The ARCSI™ (AI Regulatory Compliance Stringency Index) is a proprietary metric developed by Axis Intelligence Research to score the regulatory burden imposed by any jurisdiction’s AI governance framework. It combines six components — binding law existence, penalty scale, system scope, enforcement infrastructure, innovation exemptions, and temporal activation — into a 0–10 score. Formula: (B×2 + P×2 + S×1.5 + E×1.5 + I + T) / 14.5 × 10. Released Q2 2026, versioned as v1.0, licensed CC BY 4.0.

What is the EU AI Act’s current enforcement status in mid-2026?

As of August 2026, three EU AI Act provisions are actively enforceable: Article 5 prohibited AI practices (since February 2, 2025), GPAI model obligations (since August 2, 2025), and Article 50 transparency requirements for chatbots and AI-generated content (from August 2, 2026). High-risk Annex III system requirements were deferred from August 2, 2026 to December 2, 2027 under the EU Digital Omnibus (provisional agreement, May 7, 2026, pending formal adoption). See Axis Intelligence’s EU AI Act enforcement guide for the complete compliance roadmap.

Which Asian country has the most advanced AI regulation?

China has the most comprehensive AI-specific legal framework in Asia, with six binding regulations enacted between 2022 and 2026, scoring 7.8/10 on the ARCSI™. South Korea is second, with its AI Basic Act (in force January 22, 2026) scoring 6.3/10 — making it the second country globally with a comprehensive, risk-based AI law. Japan scores 4.0/10 with its AI Promotion Act (May 2025), which operates through voluntary commitments and name-and-shame enforcement rather than fines. Singapore scores 2.6/10, with no binding law but the world’s first agentic AI governance framework (January 2026).

What is Brazil doing on AI regulation?

Brazil’s AI Bill (PL 2338/2023) passed the Senate on December 10, 2024, per Brazilian Federal Senate records. As of July 2026, it remains under review in the Chamber of Deputies, where a vote was anticipated before the August 2026 recess per multiple legal analyses. The bill adopts a risk-based EU-inspired approach, designates the ANPD as coordinating regulator, and aligns with Brazil’s constitutional human rights framework. Axis Intelligence Research’s ARCSI™ scores Brazil at 3.3/10, reflecting the pending status — a score that would rise materially if the bill is enacted.

How does AI regulation differ between developed and developing economies?

Stanford HAI’s 2026 AI Index found that over half of newly adopted national AI strategies in 2024 came from emerging economies — but the resulting enforcement infrastructure is “wildly uneven.” The compliance cost gap between the most and least stringent jurisdictions runs eightfold. High-income countries account for 87% of notable AI model production and 91% of AI startup funding, while low-income countries control 0.1% of global data center capacity. For Africa, the OECD’s March 2026 workshop in Nairobi found that most countries still lack national AI strategies, and those that have adopted strategies face infrastructure and institutional capacity constraints that make advanced EU-style regulation impractical in the near term.

Is Canada regulating AI?

No binding AI law exists in Canada as of July 2026. The Artificial Intelligence and Data Act (AIDA), introduced in 2022, died when Parliament was prorogued on January 6, 2025, and has not been reintroduced. Canada’s current framework is the voluntary Voluntary Code of Conduct on Responsible GenAI, with approximately 40 company signatories. Axis Intelligence Research scores Canada at 1.1/10 on the ARCSI™ — the lowest score among major economies in this analysis, despite Canada being among the first countries to propose AI legislation.