Artificial IntelligenceRegulations
A worldwide reference · current to September 14, 2026

The law is catching up
to artificial intelligence.

Every major jurisdiction, cited to primary text, kept current.

Open the register Browse by topic
33national and EU regimes
11US states tracked separately
3comprehensive AI laws in force
7%of global turnover, the top EU fine
The register

Every regime, at a glance

Search and filter all →

National and EU regimes, with status reflecting binding law rather than proposals or guidance. The United States sits as one entry, because its federal posture and its state laws are two different questions. Select any row for the governing laws, timeline, penalties, and primary sources.

FlagJurisdictionRegulatory approachStatus
ChinaExtensive binding sectoral rules; no single comprehensive actIn forceEuropean UnionComprehensive, horizontal, risk-based statutePhasing inSouth KoreaComprehensive risk-based act with a light enforcement postureIn forceUnited KingdomSectoral, principles-based; no horizontal AI act by designNo comprehensive lawUnited StatesNo comprehensive federal law; active state legislation and a preemption fightNo comprehensive lawAustraliaNo AI act; existing laws plus voluntary guidance, by explicit choiceNo comprehensive lawBrazilComprehensive EU-style bill advancing through the legislatureProposedCanadaProposed federal act lapsed; voluntary code plus provincial rulesLapsed / failedCouncil of EuropeFirst binding international AI treaty (framework convention)Partially in forceJapanInnovation-first framework law; light-touch and non-punitiveIn forceMexicoNo AI law; bills pending, data protection reform added ADM rightsProposedPeruFirst AI law in Latin America; promotional statute with risk-based regulationIn forceSaudi ArabiaCentralized governance through SDAIA; principles and drafts, no AI actNo comprehensive lawSwitzerlandDeliberate counter-model to the EU: sectoral, with a treaty-implementation bill dueProposedTaiwanFramework statute; principles in force, obligations to followIn forceUAELayered model: charters and strategy federally, binding rules in zones and financePartially in forceVietnamComprehensive, risk-based AI law with extraterritorial reachIn forceArgentinaNo AI law; guidance and hearings, momentum without a vehicleNo comprehensive lawChileEU-inspired risk-based bill in Congress; updated national policyProposedColombiaThree bills in play; policy framework set, no lawProposedEgyptStrategy and charter; data protection law binds, no AI actNo comprehensive lawHong KongNo AI law; dense sectoral guidance, hardest in financeNo comprehensive lawIndiaPrinciples-based guidelines plus targeted synthetic-media rulesNo comprehensive lawIndonesiaRoadmap and ethics guidance; binding rules in developmentProposedIsraelDeliberately decentralized: sectoral regulators, soft law, no AI actNo comprehensive lawKenyaAI Bill tabled 2026: classification, high-risk duties, and a registerProposedMalaysiaVoluntary governance guidelines; no binding AI lawNo comprehensive lawNew ZealandNo AI-specific law; existing technology-neutral statutesNo comprehensive lawNigeriaBinding framework in motion: risk-based bill with licensing for high-risk AIProposedSingaporeVoluntary frameworks and testing tools; no binding AI lawNo comprehensive lawSouth AfricaDraft national AI policy; POPIA binds, no AI legislation proposed yetNo comprehensive lawThailandDraft AI law in development; data protection already appliesProposedTurkeyDraft AI framework pending; data protection and cybersecurity law bind nowProposed
In forcePartially in forcePhasing inProposedNo comprehensive lawRepealed / replacedLapsed / failed
United States, state by state. Eleven states sit on the tracker, from comprehensive frameworks to single-issue statutes to one vetoed law, and most follow the resident rather than the company. Open the state tracker to sort and filter them. State tracker →

AI outran the law for a decade. That has reversed: the European Union, South Korea, and Vietnam now have comprehensive AI statutes in force, and the United States is fighting itself over whether its eleven state laws should exist at all.

There is no global rulebook, only national bets that reach past their borders, so global products build to the strictest regime they touch, usually Brussels, and subtract. Every jurisdiction below is filed the same way and dated, with every deadline through 2028, the penalty table, every law by name, and the vocabulary one click down.

The two models

Two ways to regulate a technology

Almost every regime falls into one of two shapes, with China standing apart. Knowing which shape governs you tells you where to look first.

Model 01

Comprehensive and risk-based

One law covers the whole economy and scales its obligations to how much harm a use could cause. Banned uses at the top, heavy duties for high-risk uses, light disclosure for the rest. The template is the EU AI Act. These regimes tend to reach across borders and carry the largest fines.

European UnionSouth KoreaVietnamBrazil
Model 02

Sectoral and principles-based

No dedicated AI law. Existing regulators apply broad principles or established statutes to whatever AI does inside their remit. Lighter and more adaptable, and harder to map, because the rule that governs you depends on what the system is used for, not on the fact that it is AI.

United KingdomUnited States (Federal)JapanSingaporeIndiaAustraliaIsraelSwitzerland
A model of its own

China

China fits neither. It runs the most operationally demanding regime in the world through a stack of binding rules on recommendation algorithms, deep synthesis, generative services, and content labeling, all tied to content control and national security, and all enforced without a single comprehensive act.

China
What is coming

Key deadlines through 2028

Full calendar →

The dates that will actually change how AI is built and shipped. Most of the near-term pressure is European, and most of it lands in the second half of 2026.

  • September 30, 2026California governor's deadline on 30-plus AI bills
    Sign, veto, or let become law: SB 1000 (Transparency Act threshold removal), SB 947 (no discipline or termination decided by AI alone), AB 1609 (customer-service chatbots), AB 1979 (health-care chatbots as medical providers), AB 2713 (platform provenance data), and workplace monitoring and displacement bills.
  • October 1, 2026Connecticut's CART Act takes effect: first wave
    Employment-tool developer duties, the amendment making algorithmic tools no defense to discrimination claims, AI-related layoff disclosures, provenance and subscription rules.
  • December 2, 2026EU ban on non-consensual intimate imagery generators; watermark grace period ends
    The Digital Omnibus's Article 5 addition takes effect and the transitional accommodation on marking closes.
  • End of 2026Switzerland's AI consultation bill due
    The FDJP's draft implementing the Council of Europe convention, covering transparency, data protection, non-discrimination, and oversight.
  • January 1, 2027Colorado ADMT law, New York frontier law, and Illinois' AI Safety Measures Act, and Connecticut's AI companion rules take effect; Peru's first sector wave completes
    Colorado's replacement statute, New York's frontier transparency law, and Illinois' frontier safety act all begin; Peru's health, education, justice, security, and finance sectors reach their compliance mark.
  • October 1, 2027Connecticut employment-tool deployer duties apply
    Deployers of automated employment decision technology must give notice and meet the CART Act's transparency duties for systems deployed from this date.
Enforcement

The teeth vary wildly

All 44 compared →

Headlines fixate on the size of European fines, but the real deterrent differs by jurisdiction. In some places the risk is a fine; in others it is losing the right to operate, or a class action.

European Union

Up to 35 million euros or 7 percent of worldwide turnover for a banned use, the steepest fines in force anywhere. They bind companies based outside the Union.

China

No headline fine, but harder teeth in practice: service suspension, content takedowns, and revoked filings, with more immediate penalties after the 2026 Cybersecurity Law amendment.

South Korea

Modest by design. Administrative fines up to about 30 million won, roughly twenty thousand dollars, with a grace period through 2026 to ease companies in.

United States

No federal AI fine exists. Exposure runs through state attorneys general, the FTC's deception authority, and private suits under biometric laws such as Illinois BIPA, where damages stack per violation.

By theme

Regulation by topic

All topics →

The same problem, deepfakes or hiring tools or frontier models, is handled differently across borders. Each page lines the jurisdictions up side by side.

Prohibited AI practices

Uses of AI that are banned outright rather than merely regulated.

Compared across 5 jurisdictions →

High-risk systems

AI used in consequential decisions: hiring, credit, health, and public services.

Compared across 19 jurisdictions →

Transparency and disclosure

Telling people when they are dealing with AI or AI-generated content.

Compared across 32 jurisdictions →

Deepfakes and synthetic media

Rules on AI-generated likenesses, voices, and non-consensual imagery.

Compared across 8 jurisdictions →

Biometrics and facial recognition

Facial recognition, emotion recognition, and biometric categorization.

Compared across 7 jurisdictions →

Automated decisions and employment

AI in hiring, performance, promotion, and termination.

Compared across 13 jurisdictions →

Generative AI and foundation models

Rules aimed at general-purpose models and the companies that build them.

Compared across 10 jurisdictions →
News digest

What changed recently

Full digest →

Live headlines refresh automatically every six hours. Below them, original summaries of the moves that matter, each linked to the source.

September 10, 2026
Governor of California

Adam's Law: California mandates independent child-safety audits of companion chatbots

Signed September 10 with a twelve-bill child online-safety package, SB 1119 requires companion-chatbot operators to run crisis protocols for suicidal ideation, offer parental controls, notify parents when a child disables safety settings, and, for the first time in any jurisdiction, submit to independent child-safety audits and annual risk assessments. SB 867 bans companion chatbots in toys, other bills prohibit addictive feeds for users under 16 and extend child sexual exploitation law to AI-generated material. Thirty-plus further AI bills sit with the governor until September 30, including SB 1000, which would strip the one-million-user threshold from the AI Transparency Act.

CaliforniaHigh-risk systemsTransparency and disclosureSource ↗
September 9, 2026
Governor of California

California builds the audit layer: SB 813 and AB 1405 create the first state AI auditor framework

Signed September 9, SB 813 establishes independent verification organizations that can assess AI systems and models for compliance with state law, and AB 1405 creates a state registry of AI auditors with standards for independence, transparency, and integrity. Anthropic and OpenAI both endorsed the bills before signing, and the governor paired the signing with a call for federal regulation. The practical effect is infrastructure: Illinois' 2028 mandatory audits and California's own chatbot audits now have a defined class of auditors to draw on, and the frontier-template states have a model for verification rather than self-attestation.

CaliforniaHigh-risk systemsGenerative and foundation modelsSource ↗
September 4, 2026
The register

September briefing: one new state law, one regulation, one order, and a next deadline in 27 days

The register's monthly review. August was digestion after the August 2 double effective date, but the review turned up three things the register had wrong or missing: Connecticut's CART Act, signed in May and starting October 1, had been recorded as a stalled framework; the EU's Digital Omnibus has been a published regulation since July 27; and a June 2 executive order created the first federal frontier-model mechanism. All three are corrected on their pages. Next up: October 1 in Connecticut, December 2 in the EU, and the January 1 cluster.

United StatesHigh-risk systemsTransparency and disclosureSource ↗
August 2, 2026
European Commission

Enforcement begins: the AI Office takes over the AI Act

Beyond the transparency duties, August 2 was the day the AI Act grew teeth. The European Commission's AI Office, working with national authorities, formally began enforcing the act, and the Commission's power to fine providers of general-purpose AI models became applicable the same day. The obligations those fines attach to have been in force since August 2025; what changed is that noncompliance now has an enforcer with jurisdiction and a penalty schedule rather than a grace period.

European UnionGenerative and foundation modelsTransparency and disclosureSource ↗
How to read this

Common questions

Is there a global AI law?
No single global AI law exists, and none is close. As of August 2026 three comprehensive AI laws are in force: the European Union AI Act, South Korea's AI Basic Act, and Vietnam's AI Law. Taiwan and Peru have framework statutes, China enforces an extensive set of binding sectoral rules, the United States has no federal AI statute and regulates mainly through state law, and countries such as the United Kingdom, Japan, Australia, and Singapore rely on existing regulators and voluntary frameworks.
Which country has the strictest AI regulation?
The EU AI Act sets the strictest binding legal benchmark, with fines up to 35 million euros or 7 percent of worldwide turnover, which is why most global companies build to it. China's regime is often the most demanding to operate under, because of pre-deployment security assessment and dual visible-and-embedded content labeling.
Do US state AI laws still apply after the December 2025 federal executive order?
Yes. The federal preemption effort is ongoing and unsettled. State AI laws remain enforceable unless and until a court strikes them down or Congress passes a preempting statute. The prudent course is to keep complying with applicable state law while tracking the litigation.
Does a company outside the EU have to comply with the EU AI Act?
Often yes. The Act applies wherever the provider or deployer sits if the system's output is used in the EU. A model built and hosted elsewhere still falls in scope once its results reach users in the Union.
How current is this resource?
Each jurisdiction page carries a date it was last reviewed, and the register reflects binding law as of that review. AI law moves fast: Colorado repealed its landmark act before it took effect, and the EU deferred its high-risk deadlines by more than a year. Always confirm against the linked primary source before you rely on an entry.