Short answer. The EU AI Act is the only comprehensive, binding AI statute in force across a large market, and its high-risk obligations for Annex III systems apply from 2 August 2026. The US has no federal AI statute; Colorado, California, Texas and Illinois have state rules dated 2026. China, South Korea and Japan regulate specific uses or promote rather than prohibit.
Baseline: 20 November 2026. This is a research aid, not legal advice. Several instruments were amended or proposed in 2025–2026, and a date on this page can change before it takes effect. Verify in the official gazette and take counsel review before relying on a row for compliance.
Status key. In force is binding and applicable; phased is binding with later obligations; pending is not yet enacted; guidance is voluntary; no AI statute means privacy and sector rules still apply.
The 30 jurisdictions
| Jurisdiction | Instrument | Status | Effective | API relevance |
|---|---|---|---|---|
| European Union | AI Act, Reg. (EU) 2024/1689 | Phased | 2024-08-01; 2025-02-02; 2025-08-02; 2026-08-02; 2027-08-02 | Risk tiers, GPAI transparency, high-risk conformity |
| United States (federal) | No comprehensive statute; EO 14179 and agency rules | No AI statute | — | Sector rules, procurement, export controls |
| Colorado | SB 24-205, as delayed | In force | 2026-06-30 | Developer and deployer duties, impact assessments |
| California | SB 53, Transparency in Frontier AI Act | In force | 2026-01-01 | Frontier reporting, incident disclosure |
| Texas | HB 149, TRAIGA | In force | 2026-01-01 | Intentional use, disclosure, AG enforcement |
| Illinois | HB 3773, Human Rights Act amendment | In force | 2026-01-01 | Employment decisions, bias notice |
| New York City | Local Law 144 | In force | 2023-07-05 | Automated employment decision bias audits |
| United Kingdom | No AI Act; sector regulators and DSIT guidance | No AI statute | — | Data protection, consumer and sector duties |
| Canada | AIDA did not pass; PIPEDA and provincial law | No AI statute | — | Privacy, automated decision disclosure |
| Brazil | PL 2338 | Pending | Senate approved 2024-12-10 | Risk classification, transparency |
| Mexico | No AI statute; data protection law | No AI statute | — | Privacy, sector rules |
| Argentina | Draft AI bills; data protection | Pending | — | Privacy, consumer protection |
| Chile | AI bill (Boletín 16821-19) | Pending | — | Risk categories, explainability |
| China | Generative AI Measures; labelling rules; PIPL | In force | 2023-08-15; 2025-09-01 | Content labelling, security assessment, data localisation |
| Japan | AI Promotion Act | In force | 2025 (promulgated), phased within one year | Promotion-first; no penalties in the Act |
| South Korea | AI Basic Act | In force | 2026-01-22 | High-impact AI duties, transparency |
| Singapore | AI Governance Framework; PDPA | Guidance | 2019 onwards | Voluntary governance, privacy duties |
| India | DPDP Act 2023 and rules | Phased | Rules notified 2025; phased | Consent, breach notice, children’s data |
| Indonesia | PDP Law 27/2022; AI roadmap | No AI statute | 2024-10-17 | Privacy, electronic systems rules |
| Vietnam | AI Law | In force | 2026-03-01 (reported) | Risk management, transparency |
| Australia | No AI Act; voluntary guardrails | Guidance | 2025-09 | Privacy, consumer, sector duties |
| New Zealand | No AI Act; Algorithm Charter; Privacy Act 2020 | Guidance | 2020-12-01 | Privacy, algorithm transparency |
| Saudi Arabia | SDAIA AI Ethics Principles; PDPL | Guidance | PDPL 2024-09-14 | Privacy, ethics review |
| United Arab Emirates | No federal AI law; DIFC and ADGM data laws; AI Charter | No AI statute | — | Free-zone privacy, sector rules |
| Qatar | PDPPL 2016; AI strategy | No AI statute | 2017-06-22 | Privacy, data transfer |
| Israel | Privacy Protection Law; AI policy drafts | No AI statute | — | Privacy, database registration |
| South Africa | POPIA; draft AI policy framework | No AI statute | 2021-07-01 | Privacy, automated decision notice |
| Nigeria | NDPA 2023 | In force | 2023-06-12 | Privacy, data controller duties |
| Kenya | Data Protection Act 2019; AI strategy | No AI statute | 2019-11-08 | Privacy, automated decisions |
| Switzerland | No AI Act; revised FADP | No AI statute | 2023-09-01 | Privacy, sector regulation |
“In force” means binding, even if some obligations start later. Where a date is marked promulgated or notified, implementing rules may set a different compliance date.
What changes for API developers
- Transparency is the common floor. The EU Act, China’s labelling rules and California’s frontier law all require disclosure: that a system is AI, that content is synthetic, or that a model is frontier. A chat product with no disclosure is the easiest target.
- GPAI duties sit with the model provider. The EU’s 2 August 2025 obligations cover documentation, a training-content summary and a copyright policy. Downstream developers still owe transparency and risk duties.
- High-risk classification is use-based. The same model is high-risk in hiring and minimal-risk in a recipe app, so an API provider cannot classify it for customers.
- Employment is the busiest enforcement area. New York City, Illinois, Colorado and Texas all attach duties to AI used in hiring. If your product ranks candidates, check state rules first.
Dates to plan for
| Date | What happens | Who it affects |
|---|---|---|
| 2026-01-01 | California SB 53, Texas TRAIGA and Illinois HB 3773 take effect | Frontier developers, deployers and employers |
| 2026-01-22 | South Korea’s AI Basic Act applies | Providers and deployers serving Korea |
| 2026-03-01 | Vietnam’s AI Law takes effect | Providers serving Vietnam |
| 2026-06-30 | Colorado AI Act obligations apply | Developers and deployers of high-risk systems |
| 2026-08-02 | EU high-risk obligations for Annex III systems apply | Providers of high-risk AI in the EU |
The EU dates are the most likely to move: the Commission proposed a digital omnibus in November 2025 that would delay some high-risk obligations. Check the adopted text on EUR-Lex before planning around 2 August 2026.
Penalties
The EU Act sets the highest ceiling: up to €35 million or 7% of worldwide turnover, whichever is higher, for the most serious breaches, with lower tiers below. Most US state laws are enforced by attorneys general, with penalties per violation. China relies on content and data-security penalties; Japan’s Act contains none. The numbers are not comparable, and an injunction usually arrives before a fine.
RSS and update method
Reviewed monthly and within five business days of a gazetted change. Each row carries the instrument name, status and date so it can be found in an official source. The RSS feed publishes row-level changes, so one jurisdiction routes to one owner.
Two rules: a bill is never listed as “in force”, and a date is never taken from a press release when the gazette says otherwise.
Sources
Links checked 20 November 2026. We cite the primary page for each figure. Where a provider publishes a price behind a login, we say so and do not guess.
- EU AI Act, Regulation (EU) 2024/1689 — the enacted text, recitals and application dates
- European Commission, AI Act overview — implementation timeline and guidance
- Colorado General Assembly, SB 24-205 — the enacted state statute and its delayed effective date
FAQ
Which AI regulations are in force in 2026?
The EU AI Act applies in phases, with high-risk obligations for Annex III systems from 2 August 2026. China’s generative AI measures, Japan’s AI Promotion Act and South Korea’s AI Basic Act are in force. In the US, Colorado, California, Texas and Illinois have state rules taking effect in 2026.
Does the EU AI Act apply to an API provider outside the EU?
It can. The Act reaches providers whose output is used in the EU or who place a system on the EU market. A provider outside the EU should take counsel review on whether it is a provider, a deployer, or both.
What is the deadline for EU high-risk AI?
2 August 2026 for Annex III systems such as employment, education, credit and essential services, and 2 August 2027 for AI embedded in products covered by Annex I. The Commission proposed a delay in late 2025, so confirm the adopted text before planning.
Is there a US federal AI law?
No comprehensive federal AI statute. Federal obligations come from sector regulators, procurement rules and export controls, while Colorado, California, Texas, Illinois and New York City have enacted their own AI or automated-decision rules.
What is the maximum penalty under the EU AI Act?
Up to €35 million or 7% of worldwide annual turnover, whichever is higher, for the most serious breaches. Other obligations carry lower tiers. US state penalties are generally per violation and enforced by attorneys general.
Use the data, then test it on your own workload
Plugsky publishes its plan caps and model limits. Start free with no card, or compare the flat plans against per-token billing.
Start Free → See plans