AskAjay.ai

Regulation Tracker

17 min read · ·

How 8 Countries Regulate AI in 2026: The Executive Comparison Guide

The AI rules of the EU, the Gulf, Asia-Pacific and the US, checked against the primary texts, with what each change asks of you.

Ajay Pundhir
Ajay PundhirAI Strategist & Speaker
Share

Last verified

A month drawn as a grid of seven columns and five rows on a cream ground. Six dark blue squares sit scattered across it, and one gold square marks the first day of the second week, the date this tracker was last checked.
Tracker art: house pattern, October 2026

Key Takeaways

  • →The EU AI Act's high-risk duties now apply from 2 December 2027 for Annex III uses such as hiring and credit, and from 2 August 2028 for AI inside regulated products.
  • →From 2 December 2026 the EU bans AI-generated child sexual abuse material and non-consensual intimate imagery, and generators already on the market must mark synthetic output.
  • →South Korea's AI Basic Act has applied since 22 January 2026; its fines stop at KRW 30 million and cover three failures only.
  • →China binds AI services through specific measures on algorithms, deepfakes, generative AI and labelling, with the large fines sitting in its data and cybersecurity laws.
  • →The US has no federal AI statute. State laws carry the binding duties: California, Texas, Illinois and Utah now, with Colorado and New York from 1 January 2027.

This tracker is a dated record of the AI laws, rules and official frameworks in force or on the way in the European Union and eight countries: the UAE, Saudi Arabia, China, Singapore, Japan, South Korea, Australia and the United States. I check every row against the law itself and link to it, so you can see what applies, from when, and when I last read it.

The biggest move of 2026 came from Brussels. The Digital Omnibus took effect on 27 July and pushed the EU's high-risk deadlines back by a year or more, while adding new bans that start on 2 December 2026. Elsewhere the change came from Seoul, from Colorado and from a run of US executive orders. The dates that land between now and January are in the October Regulation Note.

India isn't in the table yet. Its rules haven't been through the same check against the official texts, and it joins when they have.

Regulation tracker for European Union, United Arab Emirates, Saudi Arabia, China, Singapore, Japan, South Korea, Australia, United States: status, key date, the date each provision was last checked against its source, and the source.
InstrumentStatusKey dateLast verifiedSource
European Union
Prohibited AI practicesEU AI Act, Art. 5In forceeur-lex.europa.eu (opens in a new tab)
AI literacyEU AI Act, Art. 4In forceeur-lex.europa.eu (opens in a new tab)
General-purpose AI (GPAI) model obligationsEU AI Act, Arts. 51-55In forceeur-lex.europa.eu (opens in a new tab)
Transparency for AI interactionEU AI Act, Art. 50(1)In forceeur-lex.europa.eu (opens in a new tab)
Marking of AI-generated / synthetic content (watermarking)EU AI Act, Art. 50(2)In forceeur-lex.europa.eu (opens in a new tab)
Ban on AI-generated CSAM and non-consensual intimate imageryEU AI Act, Art. 5(1)(ba)-(bb) (new prohibition, Digital Omnibus)Applies from 2 Dec 2026eur-lex.europa.eu (opens in a new tab)
High-risk AI, standalone (Annex III)EU AI Act, Art. 6(2) + Annex III (Arts. 9-15 obligations)Applies from 2 Dec 2027eur-lex.europa.eu (opens in a new tab)
High-risk AI, product-embedded (Annex I)EU AI Act, Art. 6(1) + Annex IApplies from 2 Aug 2028eur-lex.europa.eu (opens in a new tab)
High-risk AI as a medical device or safety component of an MDR/IVDR deviceEU AI Act, Art. 6(1) + Annex I (Section A) referencing Reg. (EU) 2017/745 (MDR) and Reg. (EU) 2017/746 (IVDR); high-risk obligations Arts. 9-15Applies from 2 Aug 2028health.ec.europa.eu (opens in a new tab)
Standalone clinical decision-support software is a regulated medical device under the MDREU MDR, Regulation (EU) 2017/745 (MDR) - Art. 2(1) 'medical device' and Art. 10 (manufacturer general obligations); Art. 61 + Annex XIV (clinical evaluation); Arts. 83-86 (post-market surveillance); Annex I (general safety and performance requirements, incl. risk management, Section 3)In forceeur-lex.europa.eu (opens in a new tab)
MDR Annex VIII Rule 11 - classification of decision-support software (Class IIa to III) and the CE-marking consequenceEU MDR, Regulation (EU) 2017/745 (MDR), Annex VIII, Chapter III, Section 6.3 - Rule 11In forcehealth.ec.europa.eu (opens in a new tab)
MDCG 2019-11 Rev.1 - software qualification/classification guidance, revised 2025 to cover AI (MDAI)EU MDR, MDCG 2019-11 Rev.1 (June 2025) - Guidance on Qualification and Classification of Software in Regulation (EU) 2017/745 (MDR) and Regulation (EU) 2017/746 (IVDR)In forcehealth.ec.europa.eu (opens in a new tab)
Processing health data as special-category data (Art. 9)EU GDPR, Regulation (EU) 2016/679 (GDPR), Art. 9In forceeur-lex.europa.eu (opens in a new tab)
Data Protection Impact Assessment for high-risk health-data AI (Art. 35)EU GDPR, Regulation (EU) 2016/679 (GDPR), Art. 35In forceeur-lex.europa.eu (opens in a new tab)
Digital operational resilience for the financial sector (ICT risk management, incident reporting, resilience testing, ICT third-party oversight)DORA, Regulation (EU) 2022/2554 (DORA) - Art. 2 (scope); Arts. 5-16 (ICT risk management); Arts. 17-23 (ICT-related incident management, classification and mandatory reporting); Arts. 24-27 (digital operational resilience testing, incl. threat-led penetration testing); Arts. 28-30 (ICT third-party risk management); Arts. 31-44 (Oversight Framework for critical ICT third-party providers); Art. 64 (application date)In forceeur-lex.europa.eu (opens in a new tab)
United Arab Emirates
Personal Data Protection Law (federal baseline)UAE PDPL, Federal Decree-Law No. 45 of 2021In forceu.ae (opens in a new tab)
Personal data processed by autonomous and semi-autonomous systems (DIFC free zone)DIFC Regulation 10, DIFC Data Protection Regulations, Reg. 10In forceassets.difc.com (opens in a new tab)
UAE AI Charter (voluntary principles)UAE AI Charter, UAE Charter for the Development and Use of AI (12 principles)Voluntaryu.ae (opens in a new tab)
Saudi Arabia
Personal Data Protection LawSaudi PDPL, Royal Decree M/19 of 2021, as amended by M/148 of 2023In forcedgp.sdaia.gov.sa (opens in a new tab)
SDAIA AI Ethics Principles (voluntary)SDAIA AI Ethics PrinciplesVoluntarydgp.sdaia.gov.sa (opens in a new tab)
SDAIA Generative AI Guidelines (voluntary)SDAIA Generative AI Guidelines (Government and Public editions)Voluntarysdaia.gov.sa (opens in a new tab)
Global AI Hub Law, statutory 'data embassy' regime (draft)Global AI Hub Law (draft)ProposedNot fixedcst.gov.sa (opens in a new tab)
China
Algorithm recommendation service managementAlgorithm Recommendation Provisions (2022)In forcecac.gov.cn (opens in a new tab)
Deep synthesis (deepfake) content managementDeep Synthesis Provisions (2023)In forcecac.gov.cn (opens in a new tab)
Generative AI services interim measuresGenerative AI Interim Measures, Interim Measures for Generative AI Services, Arts. 7, 14, 17, 21 (2023)In forcecac.gov.cn (opens in a new tab)
Labeling of AI-generated synthetic contentAI Content Labeling Measures, Measures for Labeling AI-Generated Synthetic Content + GB 45438-2025 (2025)In forcecac.gov.cn (opens in a new tab)
Amended Cybersecurity Law: AI clause and tiered penaltiesCybersecurity Law (amended), effective 1 Jan 2026In forcecac.gov.cn (opens in a new tab)
Personal Information Protection Law: penalty exposurePIPL, Art. 66In forcecac.gov.cn (opens in a new tab)
Singapore
Model AI Governance FrameworkModel AI Governance Framework (2nd ed., 2020)Voluntarypdpc.gov.sg (opens in a new tab)
Model AI Governance Framework for Generative AIModel GenAI Framework, Model AI Governance Framework for Generative AI (2024)Voluntaryaiverifyfoundation.sg (opens in a new tab)
Model AI Governance Framework for Agentic AIModel Framework for Agentic AI, Model AI Governance Framework for Agentic AI (2026)Voluntaryimda.gov.sg (opens in a new tab)
AI Verify testing framework and toolkitAI Verify (AI Verify Foundation, 2023)Voluntaryimda.gov.sg (opens in a new tab)
CSA Guidelines on Securing AI SystemsCSA Guidelines on Securing AI Systems (2024)Voluntarycsa.gov.sg (opens in a new tab)
PDPA data-protection obligationsPDPA, Personal Data Protection Act 2012In forcesso.agc.gov.sg (opens in a new tab)
Japan
AI Promotion Act: best-efforts cooperation dutiesAI Promotion Act, Art. 7 (Law No. 53 of 2025)In forcelaws.e-gov.go.jp (opens in a new tab)
South Korea
High-impact AI dutiesAI Basic Act, Arts. 33 to 35In forcelaw.go.kr (opens in a new tab)
Generative-AI transparencyAI Basic Act, Art. 31In forcelaw.go.kr (opens in a new tab)
Extraterritorial applicationAI Basic Act, Art. 4In forcelaw.go.kr (opens in a new tab)
Domestic representative requirementAI Basic Act, Art. 36In forcelaw.go.kr (opens in a new tab)
Administrative fine triggersAI Basic Act, Art. 43In forcelaw.go.kr (opens in a new tab)
Australia
AI Ethics PrinciplesAI Ethics Principles (2019)Voluntaryindustry.gov.au (opens in a new tab)
Voluntary AI Safety Standard (10 guardrails)Voluntary AI Safety Standard (2024)Voluntaryindustry.gov.au (opens in a new tab)
Automated decision-making transparency (APP 1.7)Privacy Act 1988, APP 1.7Applies from 10 Dec 2026legislation.gov.au (opens in a new tab)
Mandatory guardrails for high-risk AI (proposed, not enacted)Mandatory guardrails for high-risk AI (proposed), Mandatory guardrails for AI in high-risk settings (proposals paper, Sep 2024)ProposedNot fixedindustry.gov.au (opens in a new tab)
United States
NIST AI Risk Management Framework 1.0NIST AI RMF, NIST AI 100-1Voluntarynist.gov (opens in a new tab)
NIST Generative AI ProfileNIST AI RMF, NIST-AI-600-1Voluntarynist.gov (opens in a new tab)
Safe, Secure, and Trustworthy AI (revoked 20 Jan 2025)US federal AI executive actions, EO 14110Supersededfederalregister.gov (opens in a new tab)
Initial Rescissions of Harmful Executive Orders and Actions (revokes EO 14110)US federal AI executive actions, EO 14148In forcefederalregister.gov (opens in a new tab)
Removing Barriers to American Leadership in AIUS federal AI executive actions, EO 14179In forcefederalregister.gov (opens in a new tab)
America's AI Action PlanUS federal AI executive actions, America's AI Action Plan + 3 EOsVoluntarywhitehouse.gov (opens in a new tab)
Ensuring a National Policy Framework for AIUS federal AI executive actions, Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence" (signed Dec 11, 2025; published Federal Register Dec 16, 2025, FR doc 2025-23092)In forcefederalregister.gov (opens in a new tab)
Promoting Advanced Artificial Intelligence Innovation and SecurityUS federal AI executive actions, Executive Order 14409 (signed 2 Jun 2026; 91 FR 34565, FR Doc 2026-11415)In forcefederalregister.gov (opens in a new tab)
Colorado ADMT law (transparency and human review)Colorado AI Act, Colorado SB 26-189Applies from 1 Jan 2027leg.colorado.gov (opens in a new tab)
Colorado AI Act (2024), replaced by SB 26-189 from 1 Jan 2027Colorado AI Act, Colorado SB 24-205SupersededNot fixedleg.colorado.gov (opens in a new tab)
Conversational AI service operator requirementsColorado HB 26-1263Applies from 1 Jan 2027leg.colorado.gov (opens in a new tab)
Texas Responsible AI Governance ActTexas TRAIGA, Texas HB 149 (TRAIGA)In forcecapitol.texas.gov (opens in a new tab)
Transparency in Frontier AICalifornia AI laws, California SB 53In forceleginfo.legislature.ca.gov (opens in a new tab)
Generative AI training-data transparency (documentation due 1 Jan 2026)California AI laws, California AB 2013In forceleginfo.legislature.ca.gov (opens in a new tab)
Companion chatbotsCalifornia AI laws, California SB 243In forceleginfo.legislature.ca.gov (opens in a new tab)
AI in employment decisions (Human Rights Act amendment)Illinois HB 3773In forcehrc.illinois.gov (opens in a new tab)
Utah AI Policy ActUtah AI laws, Utah SB 149 (UAIPA)In forcele.utah.gov (opens in a new tab)
AI mental-health chatbotsUtah AI laws, Utah HB 452In forcele.utah.gov (opens in a new tab)
Responsible AI Safety and Education (RAISE) ActNew York RAISE Act, NY Gen. Bus. Law Art. 44-B (RAISE Act)Applies from 1 Jan 2027nysenate.gov (opens in a new tab)

What does the EU AI Act require after the Omnibus?

The AI Act applies in stages. Its bans on practices such as social scoring and untargeted facial-image scraping have applied since 2 February 2025, along with a duty on AI literacy. Duties for general-purpose AI models followed on 2 August 2025. A year later, on 2 August 2026, the Commission gained its power to fine those model providers, and the duty to tell people they are dealing with an AI system began.

The Omnibus changed three things. High-risk duties for Annex III uses, such as hiring and credit scoring, now start on 2 December 2027, and for AI built into Annex I products on 2 August 2028. New bans on AI-generated child sexual abuse material and non-consensual intimate imagery apply from 2 December 2026, the same day that generators already on the market must mark their synthetic output. And the AI literacy duty got softer: take measures to support it, with no required level. Breaking a ban can still cost up to EUR 35 million or 7 percent of worldwide turnover. The EU AI Act guide covers the risk tiers in depth.

Open-source releases get narrow relief. AI systems under a free and open-source licence fall outside the Act unless they are high-risk, prohibited or subject to Article 50 (Article 2(12)). Open general-purpose models skip only two documentation duties and the need for an EU representative (Articles 53(2) and 54(6)); they still need a copyright policy and a summary of their training content, and the relief never covers a model with systemic risk. The Commission's guidelines for general-purpose AI providers read a licence with a user cap, a research-only clause or a separate commercial licence as not open-source, and paid access or dual licensing removes the relief too. The Omnibus left all of this unchanged.

Do the UAE and Saudi Arabia have an AI law?

Neither has an AI statute. In the UAE the binding layer is data protection. The federal Personal Data Protection Law has been in force since 2 January 2022, but its fines wait on a separate Cabinet decision that I couldn't find published. In the DIFC free zone, Regulation 10 of the Data Protection Regulations, enacted on 1 September 2023, governs personal data processed by autonomous and semi-autonomous systems. A DIFC consultation on tightening it closed on 18 July 2026 with no new rule yet. The DIFC Commissioner has published an accreditation and certification framework for Regulation 10 and a certification guide, updated on 10 August 2026, and lists approved certification bodies, so the certification Regulation 10 requires before high-risk processing can now be obtained. The UAE Charter for the Development and Use of AI, approved on 10 June 2024, sets twelve principles and no penalties. Why the data law does the work of an AI law is the subject of the UAE PDPL analysis.

Saudi Arabia's Personal Data Protection Law has been fully enforceable since 14 September 2024. In January 2026 the Saudi Press Agency reported that SDAIA's committees had issued 48 decisions confirming violations over the past year, without amounts or company names. Fines under Article 36 reach SAR 5 million and can be doubled for a repeat violation. SDAIA's AI Ethics Principles (September 2023) and Generative AI Guidelines (January 2024) are guidance for private firms. A draft Global AI Hub Law went out for consultation in April 2025 and hasn't been enacted.

What about Qatar, Bahrain, Oman and Kuwait?

None has an AI statute in force. Qatar's personal data law, Law No. 13 of 2016, has no article on automated decisions, but the Qatar Central Bank's AI Guideline, in force since 4 September 2024, requires QCB approval before a licensed firm launches a new AI system as a provider. Bahrain's data protection law, Law No. 30 of 2018, in force since 1 August 2019, lets a person ask that decisions on their work, finances, creditworthiness, reliability or conduct not be made solely by automated processing (Article 22). A 38-article AI law approved by Bahrain's Shura Council in April 2024 has not been enacted, and the May 2025 AI policy binds government entities only.

Oman amended its data protection law by Royal Decree 68/2026, in force from 7 September 2026, adding a right to object to automated decisions with human review (Article 14). Kuwait has no general data protection law: CITRA's Decision No. 26 of 2024 covers licensed telecom and ICT providers, and its national AI strategy for 2025 to 2028 was still a draft in November 2025. The Gulf Is Where AI Governance Gets Tested at Scale covers what these rules mean for a board.

Which rules bind AI services in China?

China binds AI through a stack of specific measures from the Cyberspace Administration of China and partner agencies. The first, on recommendation algorithms, has applied since 1 March 2022 and fines up to RMB 100,000. Rules on deep synthesis (deepfakes) followed on 10 January 2023, and on generative AI services on 15 August 2023; those need a filing with the algorithm registry, plus a security assessment if the service can sway public opinion. Since 1 September 2025 AI-generated content must carry labels, visible and machine-readable, under the mandatory standard GB 45438-2025.

The money sits elsewhere. In serious cases the Personal Information Protection Law reaches RMB 50 million or 5 percent of the prior year's turnover. The amended Cybersecurity Law, in force since 1 January 2026, added a clause on AI, and its fines top out at RMB 10 million. Serving China usually needs its own compliance track.

Data rules reach model training too. The Network Data Security Management Regulations, State Council Decree No. 790, in force since 1 January 2025, require providers of generative AI services to secure their training data and how it is processed (Article 19). Their 24-hour report (Article 10) covers security flaws that endanger national security or the public interest, not every data breach.

How does Singapore govern AI without an AI law?

With frameworks and testing tools, on top of data protection law. The Model AI Governance Framework (second edition, January 2020), the framework for generative AI (May 2024) and a framework for agentic AI launched on 22 January 2026 are all voluntary. So is AI Verify, the government-built testing toolkit, open-sourced in June 2023. The Cyber Security Agency's guidelines on securing AI systems (October 2024) gained an addendum on agentic AI in June 2026, which states that it is not mandatory. What binds is the Personal Data Protection Act, whose obligations have applied since 2 July 2014, with a higher penalty cap since 1 October 2022.

How much does Japan's AI Promotion Act bind?

Very little that a regulator can enforce. Law No. 53 of 2025 took effect on 4 June 2025 and in full on 1 September 2025. Article 7 asks businesses to work at using AI well and to cooperate with government measures. The Act has no penalties; the state's tools are guidance, advice and information. The Cabinet adopted a new AI Basic Plan under it on 14 July 2026. Liability still runs through existing law: copyright, personal information and the civil and criminal codes.

What does South Korea's AI Basic Act require?

The Framework Act on the Development of AI has applied since 22 January 2026, with an amendment phased in by 21 July 2026. Providers of generative or high-impact AI must give advance notice that AI is in use and label its output (Article 31). High-impact AI, in areas such as healthcare, hiring, lending and public decisions, carries risk management, explanation and human oversight duties (Article 34), plus a best-efforts impact assessment that must now reflect the needs of vulnerable groups (Article 35). Foreign operators above set thresholds must appoint a local representative (Article 36), and the Act reaches conduct abroad that affects Korean users (Article 4).

By EU standards the penalties are light. The ceiling is KRW 30 million, and only three failures trigger a fine: no advance notice, no local representative, and ignoring a stop or corrective order. The science ministry has promised at least a year of grace from January 2026.

Has Australia passed an AI law?

No. Australia proposed mandatory guardrails for high-risk AI in September 2024 and did not take them forward. Its National AI Plan of 2 December 2025 builds on existing laws and regulators and set up an AI Safety Institute. The AI Ethics Principles (2019) and the Voluntary AI Safety Standard (2024) now sit inside the Guidance for AI Adoption of October 2025, all voluntary.

The binding change is in privacy law. From 10 December 2026, APP 1.7 requires privacy policies to say when automated decisions using personal information significantly affect people, with a civil penalty for a breach. The Robodebt analysis explains why that duty lands where it does.

Sector regulators already apply existing rules to AI. For banks, insurers and super funds, APRA's information security standard CPS 234 has applied since 1 July 2019 and its operational risk standard CPS 230 since 1 July 2025, with changes for exempt service providers from 1 July 2026. APRA's letter to industry on AI of 30 April 2026 warned that governance, risk management, assurance and operational resilience practices are not keeping pace with AI adoption. ASIC's review of 23 licensees (29 October 2024) counted 624 AI use cases and found governance that in some cases lagged the adoption of AI.

Elsewhere, the OAIC has published two guidance notes, on using commercially available AI products and on developing and training generative AI models. The TGA regulates AI in medical devices and published its review of that framework in July 2025, with 14 key findings. eSafety won a A$343,500 Federal Court penalty in September 2025 against a man who posted deepfake images of Australian women. Commonwealth agencies follow the DTA's policy for the responsible use of AI in government, version 2.0, effective 15 December 2025, which requires accountable officials, transparency statements, use case registers and impact assessments. A private member's Automated Decision-Making (Safeguards and Transparency) Bill 2026 was introduced on 7 September 2026 and has not passed.

Is there a federal AI law in the United States?

No. Federal policy runs through executive orders that bind agencies, not companies. Executive Order 14110 was revoked on 20 January 2025. EO 14179 ordered an action plan, released in July 2025. EO 14365 (December 2025) directs agencies to review and challenge state AI laws but preempts none by itself. EO 14409 (June 2026) offers frontier developers a voluntary route to give the government early access to new models and rules out licensing. The NIST AI Risk Management Framework stays voluntary.

Export controls are where federal rules have touched model weights. The January 2025 AI Diffusion rule required a licence to export the weights of certain advanced closed-weight models and left open-weight models out. Commerce announced its rescission on 13 May 2025 and told enforcement staff not to enforce it. The formal rescission was still listed as a planned rule in the August 2026 regulatory agenda.

The binding rules are in the states. California's frontier-model law, SB 53, and its companion-chatbot law took effect on 1 January 2026, as did Texas's Responsible AI Governance Act and Illinois's ban on discriminatory AI in employment decisions. Utah's AI Policy Act runs until 1 July 2027. Three more arrive on 1 January 2027: Colorado's automated-decision law, SB 26-189, which replaces its 2024 AI Act; Colorado's chatbot law, HB 26-1263; and New York's RAISE Act for frontier developers. Colorado's Attorney General is taking comments on draft rules until 26 October 2026.

Which soft-law texts sit behind these laws?

Two global texts bind no one and still shape the table. The OECD AI Principles, a Council Recommendation of 22 May 2019 revised on 3 May 2024, set five principles for trustworthy AI. Fifty-one adherents back them, including the EU and every country above except the UAE and China, and the EU AI Act defines an AI system in words that closely follow the OECD's. UNESCO's Recommendation on the Ethics of AI, adopted on 23 November 2021, is the other. The OECD explainer covers what the principles ask, and the UNESCO piece covers the ethics recommendation.

What should a multinational do with this table?

Build one governance core to the strictest regime you sell into, then add a module per market. For most firms the EU sets that core: risk classification, documentation, human oversight, transparency and incident handling. Lighter regimes then become subtractions rather than new projects. China is the exception, because its filing, labelling and content rules need a track of their own. The US state laws mostly ask for disclosure and a route to human review, which an EU-grade core already produces. Minimum viable governance is the ninety-day version of that core.

Two habits matter more than the model. Give each jurisdiction one owner who reads the primary text when it moves. And plan against dates checked at the source, because they move: the EU's high-risk deadlines, New York's RAISE Act and Colorado's whole regime all changed in 2026.

What changed

  1. United States

    Corrected the RAISE Act penalty citation: the Attorney General sues under § 1427, and the state's DFS office levies the USD 1,000 daily penalty for a missed disclosure filing under § 1428(5)(a). Source: nysenate.gov (opens in a new tab)

  2. Tracker

    Added Qatar, Bahrain, Oman and Kuwait; DIFC Regulation 10 certification and Saudi PDPL enforcement; Australian sector regulators; the EU open-source relief; China's Decree 790 on training data; and the status of the US AI Diffusion rule.

  3. Tracker

    Added a section on the soft-law texts behind the table: the OECD AI Principles and UNESCO's Recommendation on the Ethics of AI.

  4. Tracker

    Rebuilt from the April 2026 eight-country guide. Every row was re-read at its primary source; the US joined the table and India left it until its rules are checked the same way. Claims that could not be traced to a primary text were cut.

  5. European Union

    The duty to tell people they are dealing with an AI system started to apply, as did the Commission's power to fine general-purpose AI model providers. Source: eur-lex.europa.eu (opens in a new tab)

  6. European Union

    The Digital Omnibus (Regulation (EU) 2026/1744) entered into force and moved the high-risk obligations for Annex III systems from 2 Aug 2026 to 2 Dec 2027. Source: eur-lex.europa.eu (opens in a new tab)

  7. European Union

    The same act moved the high-risk obligations for AI built into Annex I products from 2 Aug 2027 to 2 Aug 2028. Source: eur-lex.europa.eu (opens in a new tab)

  8. European Union

    Article 4 was replaced: providers and deployers must take measures to support AI literacy, with no duty to guarantee a set level. Source: eur-lex.europa.eu (opens in a new tab)

  9. European Union

    New prohibitions on AI-generated child sexual abuse material and non-consensual intimate imagery were added to Article 5, applying from 2 Dec 2026. Source: eur-lex.europa.eu (opens in a new tab)

  10. South Korea

    An amendment (Act No. 21311) began to require impact assessments for high-impact AI to reflect the needs of vulnerable groups. Source: law.go.kr (opens in a new tab)

  11. Japan

    The Cabinet adopted a new AI Basic Plan under the AI Promotion Act, replacing the December 2025 plan. Source: www8.cao.go.jp (opens in a new tab)

  12. United States

    Executive Order 14409 set up a voluntary pre-release access framework for frontier models and ruled out mandatory licensing. Source: federalregister.gov (opens in a new tab)

  13. United States

    Colorado enacted HB 26-1263, setting duties for operators of public conversational AI services from 1 Jan 2027. Source: leg.colorado.gov (opens in a new tab)

  14. United States

    Colorado enacted SB 26-189, replacing its 2024 AI Act from 1 Jan 2027 with a narrower transparency and human-review law. Source: leg.colorado.gov (opens in a new tab)

  15. United States

    New York signed the RAISE Act chapter amendment, setting its effective date at 1 Jan 2027. Source: nysenate.gov (opens in a new tab)

  16. Singapore

    IMDA launched a voluntary Model AI Governance Framework for Agentic AI. Source: imda.gov.sg (opens in a new tab)

  17. South Korea

    The AI Basic Act and its Enforcement Decree took effect, with a grace period of at least one year before fines. Source: law.go.kr (opens in a new tab)

  18. China

    The amended Cybersecurity Law took effect, adding an AI governance clause, with fines of up to RMB 10 million. Source: cac.gov.cn (opens in a new tab)

  19. United States

    California's frontier-model transparency law (SB 53) and companion-chatbot law (SB 243) took effect. Source: leginfo.legislature.ca.gov (opens in a new tab)

  20. United States

    Texas's Responsible AI Governance Act (HB 149) took effect. Source: capitol.texas.gov (opens in a new tab)

Questions readers ask

Which countries have a binding AI law in 2026?

Of the nine jurisdictions on this page, the EU and South Korea have comprehensive AI statutes in force. China binds AI services through a stack of specific measures. Japan has an AI statute with no penalties. In the United States the binding AI laws are state laws, in California, Colorado, Texas, Illinois, Utah and New York. The UAE, Saudi Arabia, Singapore and Australia rely on data protection law and voluntary frameworks.

When do the EU AI Act high-risk rules apply?

From 2 December 2027 for the standalone high-risk uses listed in Annex III, such as hiring and credit decisions, and from 2 August 2028 for AI built into products covered by Annex I, such as medical devices. The Digital Omnibus, Regulation (EU) 2026/1744, set those dates when it took effect on 27 July 2026.

How often is this tracker checked?

Each row carries its own review date. Rows that are provisional, recently amended or close to a deadline are re-read every 30 days, settled law every six months, and an official publication that touches a row brings its check forward. The Last verified date at the top is the most recent of those reads.


Ajay Pundhir
Ajay Pundhir

Senior AI strategist helping leaders make AI real across four continents. Forbes Technology Council member, IEEE Senior Member.

Let's Talk

Ajay's views, from fifteen-plus years in the field. Not legal or compliance advice. See full disclaimers →
Published by AI Exponent LLC

Get The Brief

Original thinking on AI strategy, governance, and durability. Every other Tuesday.