AI Policy Radar
A live reading of where AI regulation stands, where it is heading, and what comes into force next.
Next review 3 October 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. 8 tracked obligations change status within the next twelve months.
- 01 · Compliance date24 days
California — Governor's decision deadline on the 2026 AI bills. Applies in US States from 30 September 2026. Who is affected: Frontier developers, AI deployers and assurance providers operating in California
- 02 · Compliance date25 days
Connecticut SB 5 obligations start applying. Applies in US States from 1 October 2026. Who is affected: Providers of consumer AI services, employers using automated decision processes, and frontier developers with Connecticut employees
- 03 · Latest signal · G7 / OECD / UN2 September 2026
US presses a light-touch AI line at the G20 as the EU proceeds with enforcement. At the G20 digital track the United States argued against binding AI rules while the European Commission continued applying the AI Act — the clearest statement yet that the transatlantic regulatory gap is now policy, not lag.
Compliance countdown
Fixed compliance dates and obligation triggers, ticking in real time against the current date.
California — Governor's decision deadline on the 2026 AI bills
Last day for the Governor to sign or veto the AI bills passed at the close of the 2026 session, including SB 813 on independent safety verification. Bills left unsigned at the deadline follow California's default rules rather than lapsing quietly.
Connecticut SB 5 obligations start applying
Public Act 26-15 applies from this date: AI subscription disclosures, AI companion duties, automated employment-decision requirements and frontier-developer whistleblower protections.
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.
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
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 force44
- Coming into force2
- Proposed / Draft7
- Early signal2
55 of 55 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.
September 2026
US presses a light-touch AI line at the G20 as the EU proceeds with enforcement
At the G20 digital track the United States argued against binding AI rules while the European Commission continued applying the AI Act — the clearest statement yet that the transatlantic regulatory gap is now policy, not lag.
August 2026
Commission designates ChatGPT a very large online search engine under the DSA
ChatGPT becomes the first generative AI service brought inside the Digital Services Act's strictest tier, alongside Reddit and Roblox as very large online platforms. Systemic-risk assessment, mitigation, audit and data-access duties follow four months after notification.
California closes its session having sent a slate of AI bills to the Governor
The 2026 session ended on 31 August with sixteen AI bills passed, including SB 813 on independent safety verification. The Governor has until 30 September to sign or veto.
China enforces pre-development AI ethics review alongside its standards set
State media confirmed that the April 2026 ethics review measures are being applied — requiring review before AI research projects begin — with close to 200 national AI standards now in effect.
July 2026
Digital Omnibus on AI enters into force, delaying high-risk deadlines
Regulation (EU) 2026/1744 pushes high-risk AI Act obligations to Dec 2027 (Annex III) and Aug 2028 (Annex I) and simplifies documentation duties — the EU's first formal AI Act correction.
China releases international AI ethics governance action plan at WAIC 2026
Building on its 2025 Global AI Governance Action Plan and the proposed World AI Cooperation Organization, China extends its bid to shape multilateral AI ethics norms.
UN Global Dialogue on AI Governance launches in Geneva
The first universal-membership AI governance forum convenes, alongside the Independent International Scientific Panel established by the August 2025 GA resolution.
June 2026
Colorado AI Act takes effect after delayed start
The first comprehensive US state AI law is now operative — as DOJ's AI Litigation Task Force weighs challenges under the December 2025 preemption executive order.
Norway prepares national AI Act transposition ahead of EEA incorporation
Norwegian ministries continued preparatory work on the national implementation act and authority designation while the AI Act awaits incorporation into the EEA Agreement, leaving Norwegian providers exposed to EU obligations before domestic ones.
March 2026
New York RAISE Act frontier-safety duties begin
New York joins California in requiring large frontier developers to publish safety protocols and report incidents.
February 2026
India hosts the AI Impact Summit in New Delhi
The summit series' fourth leader-level meeting centers inclusive development and Global South priorities — continuing the drift from safety-first framing.
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, v2.2, 5 September 2026. https://impactlabglobal.com/policy-radar. Licensed CC BY 4.0.- 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
23 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, 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 5 September 2026
Record last verified on publication · 55 instruments · 13 deadlines · 38 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 5 September 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.