AI Policy Radar
A live reading of where AI regulation stands, where it is heading, and what comes into force next.
Next review 13 November 2026
The instrument set is reviewed on a four-week cycle. An off-cycle update is published whenever a tracked instrument is adopted, amended, struck down, or its application date moves.
What matters now
Drawn from the record rather than written by hand, so it moves as dates pass. 9 tracked obligations change status within the next twelve months.
- 01 · Compliance date15 days
New York RAISE Act — frontier developer registration opens. Applies in US States from 1 November 2026. Who is affected: Developers of frontier AI models offered in New York
- 02 · Compliance date46 days
EU AI Act — content-marking grace period ends. Applies in European Union from 2 December 2026. Who is affected: Providers of generative AI systems already on the EU market before August 2026
- 03 · Latest signal · Denmark15 October 2026
Danish Supreme Court: AI-generated child abuse material is punished like real imagery. In case 98/2025, the Højesteret held that criminal code §235 covers AI-generated depictions of minors even where the persons are fictional. It ruled that fictionality must not reduce the sentence, and upheld 18 months in prison.
Compliance countdown
Fixed compliance dates and obligation triggers, ticking in real time against the current date.
New York RAISE Act — frontier developer registration opens
Frontier AI developers must begin registering with New York State from November 2026 under the RAISE Act, ahead of full compliance duties, including 72-hour safety-incident reporting, from January 2027. A new state digital office oversees compliance.
EU AI Act — content-marking grace period ends
End of the grace period for the machine-readable marking obligation for AI systems placed on the market before 2 August 2026. From this date, all covered generative AI systems in the EU must mark outputs as AI-generated.
EU revised Product Liability Directive applies to AI products
Directive (EU) 2024/2853 applies to products placed on the market from this date — explicitly covering software and AI systems, with disclosure-of-evidence tools and eased burden of proof for claimants.
EU DSA duties bite for ChatGPT as a designated search engine
Four months after the 31 August 2026 designation, ChatGPT must meet the Digital Services Act's systemic-risk obligations: annual risk assessment and mitigation, independent audit, researcher data access, transparency reporting and a crisis-response mechanism.
Colorado re-enacted AI law takes effect
SB 26-189 replaces the suspended SB 24-205 with an automated decision-making technology act: duties against algorithmic discrimination, documentation and consumer notices for AI used in consequential decisions.
Norway — expected EEA incorporation decision on the AI Act
Indicative planning date for the EEA Joint Committee decision incorporating the AI Act into the EEA Agreement, after which Norwegian transposition sets the domestic application date. Treated as a planning marker, not a fixed legal deadline.
EU AI Act — legacy GPAI models must comply
General-purpose AI models placed on the market before 2 August 2025 must be brought into full compliance with the AI Act's GPAI obligations by this date.
EU AI Act — high-risk obligations apply (Annex III)
Post-Omnibus application date for standalone high-risk AI systems (employment, credit, education, essential services, law enforcement): full risk-management, data-governance, human-oversight and conformity obligations.
EU AI Act — high-risk obligations apply (Annex I, embedded)
Post-Omnibus application date for AI systems embedded in regulated products (machinery, medical devices, vehicles): AI Act high-risk duties integrate with sectoral conformity regimes.
Now in Effect
Connecticut SB 5 obligations start applying
California — Governor's decision deadline on the 2026 AI bills
EU AI Act — transparency obligations applied
South Korea AI Framework Act in effect
Texas TRAIGA in effect
Horizon radar
Concentric rings denote distance to impact, derived automatically from effective dates. The core is in force today; outer rings map emerging signals.
Each dot is one instrument, placed by policy domain around the circle and by distance to impact from the centre out. Hover or arrow to read it in the panel; select it to open the full entry.
Hover or focus any blip to read the instrument here. Select it to open the full entry with obligations and primary sources. Codes under each blip mark the jurisdiction.
- In force46
- Coming into force4
- Proposed / Draft7
- Early signal2
59 of 59 instruments
Early signals and adoptions
A chronological record of significant policy events. These are the events that populate the outer rings of the radar above.
October 2026
Danish Supreme Court: AI-generated child abuse material is punished like real imagery
In case 98/2025, the Højesteret held that criminal code §235 covers AI-generated depictions of minors even where the persons are fictional. It ruled that fictionality must not reduce the sentence, and upheld 18 months in prison.
Commission confirms first AI Act enforcement actions against frontier labs
EVP Henna Virkkunen said the AI Office has sent its first formal requests for information to providers of the most advanced models, covering security, external evaluation and usage monitoring, after summer incidents involving AI agents. The Commission's spokesperson had earlier confirmed RFIs to more than 30 AI companies.
UK officials say AI legislation and AISI statutory powers remain undecided
After a 12 October Commons statement on frontier and agentic AI risk, officials told MPs that ministers had not yet decided whether to legislate on AI, give the AI Security Institute statutory testing powers, or add AI safeguards to the Cyber Security and Resilience Bill.
UAE unveils an 'AI Control Tower' to monitor government AI agents
A new oversight layer will track accountability for agentic AI deployed across federal services. It follows the June 2026 AI and Data Authority consolidation. Its legal mandate is not yet confirmed in an official release (news-sourced).
Australia moves toward mandatory AI standards with incident reporting
After a consultation on national AI and data-centre standards closed, the government signalled legislation in 2027 including reporting of rogue-agent incidents. News-sourced; awaiting an official paper.
SDAIA reopens consultation on PDPL implementing-regulation changes
A revised draft adds in-Kingdom data localisation, wider breach-notification triggers and expanded registration duties. Comments close 5 November 2026.
Connecticut's AI and online safety law takes effect
Public Act 26-15 and companion privacy amendments took effect, covering AI chatbot disclosures, facial recognition and surveillance pricing.
September 2026
California enacts SB 574 on lawyers' and arbitrators' use of generative AI
The first state statute on attorney AI use. It bars delegating the practice of law to generative AI and requires lawyers to verify citations and disclose AI use in court filings. Arbitrators may not delegate decisions to AI. It takes effect 1 January 2027.
US executive order renames AI as 'Super Intelligence' in federal communications
Executive Order 14434 directs agencies to use the term Super Intelligence and asks the science office to propose a federal definition. Terminological rather than regulatory, but a definition could shape later rules.
New York sets frontier developer registration for November
Governor Hochul announced that frontier AI developers must register with the state from November 2026 under the RAISE Act, with full duties from January 2027.
Governance map
From what is coming to where it applies. Every tracked jurisdiction is shaded by the mean of its five maturity scores. Select a territory to read its regime, scoring notes and tracked instruments.
Shading is the mean of five governance dimensions (Comprehensive law, Sector & use-case rules, Institutions & capacity, Enforcement activity, International engagement). Scroll to zoom, drag to pan, select a country to open its record.
Jurisdiction maturity
The same jurisdictions scored in detail: the intensity and comprehensiveness of each AI governance regime across five dimensions.
How the radar is built
Every instrument is placed by stage and effective date rather than by editorial judgement, so the rings roll forward on their own as dates pass. Sources are official texts and primary regulator publications, linked on each instrument.
A daily scan reads official regulator and government feeds across the tracked jurisdictions and files anything that looks like a governance development into an editorial review queue. Nothing reaches the radar without a human check against the primary text, so the record stays current without becoming automated guesswork.
Data and methodology
The instrument set is published as a versioned, machine-readable dataset. Countdowns on this page are recalculated live; the records below are reviewed on a fixed cycle.
The AI Policy Radar dataset is published under Creative Commons Attribution 4.0 (CC BY 4.0). You may reuse, adapt and republish it, including commercially, provided you attribute ImpactLab as below.
ImpactLab AI Policy Radar, v2.4, 16 October 2026. https://impactlabglobal.com/policy-radar. Licensed CC BY 4.0.- v2.4 — 16 October 2026
- Full audit against primary sources across all jurisdictions; every country score re-reviewed on 16 October 2026.
- Corrected Colorado: enforcement was suspended in court on 27 April 2026 and the law was re-enacted as SB 26-189, effective 1 January 2027.
- Corrected the New York RAISE Act effective date to 1 January 2027, and Finland's supervision act to Act 1377/2025, in force 1 January 2026.
- Corrected Sweden (PTS proposed as lead authority, IMY mandated), Norway (Nkom as coordinating authority) and Denmark (three designated authorities).
- Added EU ratification of the Council of Europe AI Convention, final adoption of the AI Digital Omnibus and its new deepfake prohibition, China's AI-law commitment and Cybersecurity Law AI article, Korea's enforcement decree and fine grace period, Japan's revised AI Basic Plan, India's synthetic-content labelling rules, Singapore's agentic AI framework, Canada's AI for All strategy, the UN Global Dialogue's first session, UK legislative indecision, Saudi PDPL consultation, the UAE AI Control Tower, South Africa's draft AI policy, the Commission's first AI Act enforcement requests to frontier labs, California SB 574 on lawyers' AI use and the Danish Supreme Court ruling on AI-generated abuse material.
- Score changes: EU enforcement 2 to 3 (AI Act enforcement live since 2 August 2026); India sector rules 2 to 3 (binding synthetic-content rules).
- v2.3 — 16 October 2026
- California SB 813 moved from bill to enacted law (signed 9 September with AB 1405 on AI auditors).
- Added New York RAISE Act registration, opening November 2026.
- Added Connecticut SB 5 taking effect on 1 October, China AI Safety Governance Framework 3.0, US Executive Order 14434, the UK human rights committee call for an AI Bill, Brazil's delayed AI bill vote and Australia's move toward mandatory standards.
- Added Saudi Arabia (PDPL, National AI Risk Management Framework) and Qatar (draft National AI Policy), and updated the UAE with its new federal AI and Data Authority; the three now form a Middle East region.
- v2.2 — 5 September 2026
- Added the Commission's 31 August designation of ChatGPT as a very large online search engine under the DSA, with the four-month compliance date.
- Added Connecticut SB 5 (Public Act 26-15) and its 1 October application date.
- Added California SB 813 as passed by the legislature, with the 30 September signing deadline.
- Added China's enforcement of pre-development AI ethics review and the G20 transatlantic split.
- Rebuilt the automated source scan: dead regulator feeds replaced, and a partial scan now reports itself as degraded.
- v2.1 — 13 August 2026
- Added instrument permalinks and jurisdiction hubs with stable slugs.
- Published the machine-readable JSON and CSV exports under CC BY 4.0.
- Re-verified every EU AI Act application date against the 2026 Digital Omnibus.
- v2.0 — 18 June 2026
- Extended coverage to the full Nordic set with maturity scoring.
- Added the signals timeline as a tracked series alongside instruments.
- v1.0 — 2 March 2026
- First public release of the AI Policy Radar instrument set.
A note on “Context”: where a signal or compliance card shows a Context link, it opens the primary legal instrument or the official announcement it derives from. Where no such source is published, the instrument is named in plain text instead of an inert link.
Scoring rubric
Each jurisdiction is scored 0 to 4 on five dimensions. The same scale is applied to every dimension, and every score carries a written note in the jurisdiction panel.
- 0None / not started
- 1Early exploration
- 2Frameworks in development
- 3Rules adopted, ramping up
- 4Comprehensive & operating
- Comprehensive law
- Sector & use-case rules
- Institutions & capacity
- Enforcement activity
- International engagement
What counts, and what does not
An instrument enters the record when it is a law, regulation, bill, treaty, standard, official guidance, enforcement action or authority designation that governs the development or deployment of AI, and when a primary source text can be linked.
Excluded: funding announcements, conferences, consultations without a published position, commentary, and corporate policy. A signal enters the record when a public authority takes a step that changes the probability or shape of a future instrument.
Coverage
25 jurisdictions and blocs are tracked: European Union, United States (Federal), US States, United Kingdom, China, Canada, Japan, South Korea, Singapore, India, Brazil, Australia, United Arab Emirates, Saudi Arabia, Qatar, Italy, Norway, Sweden, Denmark, Finland, Iceland, Nordic Cooperation, Council of Europe, G7 / OECD / UN, African Union. Absence from this list means the jurisdiction is not yet tracked, not that it has no AI regulation. Coverage is deepest in Europe, North America and the Nordics, and thinner in Africa, South America and Central Asia; that gap is a limitation of the record, and additions are welcome.
Retrieved 16 October 2026
Record last verified on publication · 59 instruments · 14 deadlines · 66 signals
Take the data
The record is open. Export it, check it against the primary sources, and tell us where we are wrong.
Citation
ImpactLab. AI Policy Radar. Oslo: ImpactLab, 2026. Retrieved 16 October 2026. https://impactlabglobal.com/policy-radar
Corrections
Send a correction with the primary source to hello@impactlabglobal.com. Accepted corrections are applied to the record and the affected entry is re-dated, so the change is visible in the verification date shown on the instrument.