Artificial Intelligence Policy Tracker

AI policy updates

AI policy updates, September 2026

Summary

September 2026: 25 AI policy changes recorded across 2 jurisdictions, of which 1 concerned binding law, 4 were rated high impact. Each is dated, links to its official source, and shows what it means in practice.

Changes
25
Jurisdictions
2
Entered into force
0
Urgent or high impact
4

Last updated · most recent change 30 Sep 2026

All changes (25)

Template v3 released: Colorado ADMT Law (SB 26-189) Notices

Dataset 3335c39da4b7 → 051ce01569fc. Sheet "Checklist": 0 → 5 rows (+5). Sheet "Dates": 0 → 2 rows (+2). Document: 5 → 29 blocks. Record citations: 0 → 12.

Practical impact: Organisations using the v3 file should download the new version; the README page in each file states which dataset it was built from.

Colorado ADMT Law (SB 26-189) Notices v3 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: Global AI Regulatory Applicability Matrix

Dataset 914895c3103e → 3335c39da4b7. Sheet "Deadlines": 22 → 23 rows (+1). Record citations: 64 → 65.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

Global AI Regulatory Applicability Matrix v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: AI System Technical Documentation

Dataset 914895c3103e → 3335c39da4b7. Sheet "Documentation duties": 10 → 11 rows (+1). Document: 87 → 85 blocks. Record citations: 20 → 21.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

AI System Technical Documentation v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: Colorado ADMT Law (SB 26-189) Notices

Dataset 914895c3103e → 3335c39da4b7. Sheet "Checklist": 10 → 0 rows (-10). Sheet "Dates": 3 → 0 rows (-3). Document: 103 → 5 blocks. Record citations: 31 → 0.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

Colorado AI Act Notices v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: NIST AI RMF ↔ EU AI Act ↔ ISO/IEC 42001 Crosswalk

Dataset 7c0835db52c1 → 3335c39da4b7. Sheet "Crosswalk": 108 → 98 rows (-10). Sheet "Coverage": 17 → 16 rows (-1). Record citations: 108 → 98.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

NIST AI RMF ↔ EU AI Act ↔ ISO/IEC 42001 Crosswalk v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: ISO/IEC 42001 Gap Assessment and Statement of Applicability

Dataset 914895c3103e → 3335c39da4b7.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

ISO/IEC 42001 Gap Assessment and Statement of Applicability v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: AI Agent Registry and Permission Matrix

Dataset 914895c3103e → 3335c39da4b7. Sheet "Oversight duties": 34 → 35 rows (+1). Document: 53 → 52 blocks. Record citations: 42 → 43.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

AI Agent Registry and Permission Matrix v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: AI Incident Response Playbook and Log

Dataset 914895c3103e → 3335c39da4b7. Sheet "Incident duties": 9 → 7 rows (-2). Document: 69 → 57 blocks. Record citations: 256 → 252.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

AI Incident Response Playbook and Log v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: Human Oversight Procedure

Dataset 914895c3103e → 3335c39da4b7. Sheet "Oversight duties": 8 → 9 rows (+1). Document: 60 → 65 blocks. Record citations: 16 → 18.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

Human Oversight Procedure v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: AI Impact Assessment

Dataset 914895c3103e → 3335c39da4b7. Sheet "Duties by jurisdiction": 7 → 6 rows (-1). Document: 54 → 48 blocks. Record citations: 14 → 12.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

AI Impact Assessment v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: Fundamental Rights Impact Assessment (FRIA)

Dataset 914895c3103e → 3335c39da4b7. Sheet "Impact-assessment duties": 7 → 6 rows (-1). Record citations: 8 → 7.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

Fundamental Rights Impact Assessment (FRIA) v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: EU AI Act Role and Risk Classifier

Dataset 7c0835db52c1 → 3335c39da4b7. Sheet "Dates": 7 → 9 rows (+2). Record citations: 51 → 53.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

EU AI Act Role and Risk Classifier v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: AI Risk Register (MIT AI Risk Repository taxonomy)

Dataset 914895c3103e → 3335c39da4b7. Sheet "Risk duties": 13 → 10 rows (-3). Record citations: 39 → 36.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

AI Risk Register (MIT AI Risk Repository taxonomy) v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)

Template v2 released: AI System Inventory

Dataset 914895c3103e → 3335c39da4b7. Sheet "Duties by role": 112 → 107 rows (-5). Document: 77 → 75 blocks. Record citations: 166 → 161.

Practical impact: Organisations using the v2 file should download the new version; the README page in each file states which dataset it was built from.

AI System Inventory v2 Verified (reviewed by templates:build; a factual check against the official source, not a legal review)
California (United States) Routine

California executive order directs work on AI oversight, including a frontier-model kill switch

The Governor issued an executive order asking a working group to recommend, within about two months, measures to strengthen California's AI safety and security laws, including whether developers of frontier models should maintain a shutdown capability and whether independent third parties should prepare safety plans. It creates no new obligations itself. Recorded from secondary reporting; confirm the order's number and text on the Governor's site.

Practical impact: No obligations yet; frontier-model developers should watch for the working group's recommendations.

Governor Newsom issues executive order to accelerate independent oversight and advance the creation of an AI kill switch Verified (reviewed by Bhaskar Bhatt; a factual check against the official source, not a legal review)
California (United States) High Adopted

California SB 1050 requires disclosure of AI synthetic performers in advertising

SB 1050 makes it unlawful to create or publish an advertisement that prominently includes a synthetic performer (a human-like digital voice, figure or representation created at least partly with generative AI and not depicting an identifiable person) without a clear and conspicuous disclosure. It takes effect on 1 January 2027. Recorded from secondary reporting; confirm against the chaptered text.

Practical impact: Advertisers and agencies should add a clear disclosure to any California advertisement that features an AI-generated performer from 1 January 2027.

SB-1050 False advertising: synthetic performers Verified (reviewed by Bhaskar Bhatt; a factual check against the official source, not a legal review)
California (United States) High Adopted

California signs SB 1119 on companion chatbots and children's safety

SB 1119 was chaptered (Chapter 190, Statutes of 2026). Operators must carry out a child-focused risk assessment before releasing a new or substantially modified companion chatbot from 1 July 2027, and commission child-safety audits by 1 January 2029. Recorded from search extracts of the legislature's pages.

Practical impact: Companion chatbot operators should plan a documented child-safety risk assessment before mid-2027 releases.

California chatbot child-safety law (SB 1119) California Legislative Information: SB 1119 (2025-2026) status Pending review (a factual check against the official source, not a legal review)
California (United States) High Adopted

California signs child-safety laws covering companion chatbots and social media

The Governor signed a package of child-safety bills that, among other things, require risk assessments before new companion-chatbot rollouts, add penalties for harm to children and restrict toys that include a companion chatbot. According to the announcement and secondary reporting, most provisions take effect on 1 January 2027 and the companion-chatbot safeguards on 1 July 2027. Recorded from secondary reporting; confirm each bill against its enrolled text.

Practical impact: Providers of companion chatbots available to minors in California should plan risk assessments and safeguards for the 2027 dates.

Governor Newsom signs the strongest child safety chatbot and social media laws in the nation Verified (reviewed by Bhaskar Bhatt; a factual check against the official source, not a legal review)
California (United States) High Adopted

California enacts an AI audit framework with SB 813 and AB 1405

The Governor signed SB 813, under which the Government Operations Agency sets criteria by 1 January 2028 for designating qualified AI auditors as independent verification organisations, and AB 1405, which creates an online AI Auditor Registry by 1 January 2029, after which a covered AI audit may be offered only by a registered auditor meeting independence and integrity standards. Recorded from secondary reporting; confirm against the enrolled bills.

Practical impact: Organisations that commission AI audits for California compliance should plan to use auditors that register under AB 1405 once the registry opens.

Governor Newsom signs first-in-the-nation AI safeguards to protect Californians Verified (reviewed by Bhaskar Bhatt; a factual check against the official source, not a legal review)

Disclaimer: informational only, not legal advice. Verify every claim against the linked official sources and consult a qualified lawyer before acting.