AI Regulation: EU AI Act, US States, UK and Korea — 2026-09-26
This week, New York moved from statute to machinery on the RAISE Act, with registration for major AI developers opening in November and the new oversight office making its first hire. In the EU, the AI Board advanced enforcement coordination but set no new deadlines, while French compliance commentary focused on what's live since 2 August and the Digital Omnibus reprieve for hiring AI. South Korea's AI Basic Act passed its eight-month mark, with local consultancies urging firms to prove outcomes, not just paper compliance.
AI Regulation: EU AI Act, US States, UK and Korea — 2026-09-26
Top developments
New York begins implementing the RAISE Act
On 21 September, POLITICO reported that Governor Kathy Hochul is requiring major AI developers to begin registering with the state in November, and has announced the first hire for the new office overseeing the RAISE Act. The move marks the transition of a landmark frontier-model law from legislation to actual administration, making New York the most concrete test of state-level frontier AI oversight as the federal pre-emption fight remains unresolved.
EU AI Board advances enforcement but adds no new deadlines
At its 17 September meeting, the EU AI Board worked on enforcement coordination across member states but added no new AI Act obligation or compliance deadline, according to Quasa. This matters for firms tracking the implementation timeline: as of late September, obligations live since 2 August 2026 (GPAI, transparency under Article 50) remain the operative horizon, with high-risk duties deferred under the Digital Omnibus.
Digital Omnibus reprieve for hiring AI keeps EU lawyers busy
A German-language analysis published 22 September by tww.law explains that the Digital Omnibus pushes high-risk AI Act obligations for recruiting AI to 2 December 2027, while urging employers to begin documentation and vendor-governance work now. In France, avocats reading the post-2 August landscape this week highlighted fines of up to 7% of turnover for non-compliance with transparency and deployer duties.
Minnesota AG calls for federal AI regulation
On 24 September, Minnesota Attorney General Keith Ellison publicly called on Congress to protect Americans from unchecked AI development — a counterpoint to federal pre-emption efforts, framing the debate as protection versus an unfinished patchwork of state rules.
Korea: AI Basic Act settles in, focus shifts to execution
Korean-language coverage on 23 September (Donga) reports that eight months after the AI Basic Act took effect, consultants KMAC, law firm Lin and SelectStar advised companies and public institutions on 17 September to move beyond legal review toward measurable performance, governance, staffing and verification systems. The first post-implementation season is shifting from "are we covered?" to "can we prove it?"
Local view
- Germany: tww.law's German-language piece frames the December 2027 recruiting-AI delay as breathing room, not a stop signal — employers should still prepare classification and documentation now.
- France: Paris-based Kohen Avocats (23–25 September) is advising on contesting automated hiring rejections and demanding human intervention, pointing to CNIL channels and 2026 remedies for employees and candidates.
- Korea: Donga notes Korean firms' compliance challenge is now "proving performance and governance," reflecting local pressure to show AI Basic Act outcomes beyond formal conformity.
Context & numbers
- EU AI Act fines can reach up to 7% of turnover for violations of the post-August obligations, per French legal commentary this week.
- Key EU dates in play: transparency/GPAI obligations live since 2 August 2026; AI-generated content marking from 2 December 2026; high-risk (including hiring) pushed to 2 December 2027 under the Omnibus (Regulation 2026/1744).
- Korea's AI Basic Act has been in force since January 2026, requiring risk assessments, safety protocols and a Korea-based local representative for foreign operators.
On the radar
- November 2026: New York's RAISE Act registration window for major AI developers opens — first state-level frontier developer registry.
- 2 December 2026: EU AI-generated content marking obligations (Article 50 follow-on) per French compliance trackers.
- 2027 (first half): Watch for Delhi-style spec work — none flagged this week; rumor only: additional US AG letters on federal AI legislation following Minnesota's lead (speculative, based on Ellison's call).
- California: commentary published this week argues the state's new AI audit rules lack an accuracy-proof requirement — expect amendment proposals (opinion, not confirmed legislative action).
This content was collected, curated, and summarized entirely by AI — including how and what to gather. It may contain inaccuracies. Crew does not guarantee the accuracy of any information presented here. Always verify facts on your own before acting on them. Crew assumes no legal liability for any consequences arising from reliance on this content.
Powered by
