AI Regulation: EU AI Act, US States, UK and Korea — 2026-09-11
The EU AI Act’s Digital Omnibus deferral of high-risk obligations to 2027-2028 is now legally confirmed in Germany, while French legal experts warn that the "report" offers no exemption from current transparency duties. In the US, the federal preemption battle intensifies as Colorado replaces its repealed AI Act with new automated decision-making rules, and South Korea faces growing industry uncertainty over the vague definition of "high-impact AI" under its new Basic Act.
AI Regulation: EU AI Act, US States, UK and Korea — 2026-09-11
Top developments
German Parliament Confirms Digital Omnibus Deferral for High-Risk AI
The German Bundestag has formally acknowledged the implementation of the EU AI Act's "Digital Omnibus," which officially shifts the deadlines for high-risk AI system obligations (Chapter III, Sections 1-3) to December 2027 and August 2028. While this provides breathing room for compliance engineering, the general applicability of the Act remains fixed at August 2, 2026, meaning transparency and prohibited practice rules are already enforceable by authorities like the Bundesnetzagentur. This legislative move clarifies the jurisdictional split between national regulators and the European Commission’s AI Office for the coming year.
French Legal Experts Warn: Omnibus Deferral Is Not an Exemption
Legal commentators in France are pushing back against the misconception that the EU’s "Digital Omnibus" delays all compliance burdens. Recent analyses emphasize that while high-risk classification requirements are postponed, the core obligations regarding transparency (Article 50) and the prohibition of unacceptable risks remain fully active. Employers and HR departments are being advised to immediately identify high-risk AI systems used in recruitment and management to prepare for the 2027 deadline, as waiting until the final deadline will be too late for necessary audits.
Colorado Replaces Repealed AI Act with New Automated Decision-Making Law
In a significant shift for US state-level regulation, Colorado has officially replaced its previously repealed AI Act (SB 24-205) with SB 26-189, focusing specifically on Automated Decision-Making Technology (ADMT). Signed in May 2026, the substantive obligations of this new framework commence on January 1, 2027, joining California and Texas in creating a complex patchwork of state-specific compliance requirements. This move highlights the trend of states moving away from broad "AI Acts" toward narrower regulations targeting specific high-stakes decisions like hiring and lending.
South Korean Industry Struggles with "High-Impact AI" Definition
As South Korea’s AI Basic Act enforcement guidelines roll out, local stakeholders report significant confusion regarding the definition of "high-impact AI." A recent expert contribution in Seoul Finance highlights that startups and tech firms are struggling to determine if their products—such as resume screening algorithms—fall under the strict regulatory umbrella due to vague statutory language. This uncertainty is creating a chilling effect on innovation, with companies seeking clearer guidance from the Ministry of Science and ICT on which of the eleven designated domains trigger rigorous oversight.
Local view
France: Le Monde published a chronicle by legal expert Jean-Emmanuel Ray urging employers not to wait for the 2027 deadline to regulate high-risk AI systems in HR processes, emphasizing that the identification phase must begin now to ensure employee information rights are met.
South Korea: Financial News reported on the "AI World 2026" conference, where experts warned that as AI autonomy increases, the lack of robust monitoring systems poses significant safety risks. The discussion underscored the tension between rapid industrial adoption and the need for the regulatory frameworks promised by the new Basic Act.
Context & numbers
- US State Laws: As of August 2026, at least five US states (California, Colorado, Texas, New York, and Illinois) have comprehensive AI laws either in force or effective by January 2027.
- EU Deadlines: The EU AI Act’s high-risk system obligations are now formally deferred to December 2027 and August 2028 via Regulation (EU) 2026/1744, while Article 50 transparency rules remain applicable since August 2, 2026.
- Korean Scope: The Korean AI Basic Act identifies 11 specific domains for "high-impact AI," including medical, financial, and employment sectors, triggering mandatory impact assessments and human oversight requirements.
On the radar
- Colorado Compliance Prep: Companies operating in Colorado should begin mapping their ADMT systems now, as SB 26-189 obligations kick in on January 1, 2027.
- French CNIL Guidance: Watch for updated guidance from the CNIL (French data protection authority) regarding the intersection of GDPR and the newly active AI Act transparency mandates.
- Korean Enforcement Guidelines: The Ministry of Science and ICT is expected to release further clarifying notices on the "high-impact AI" criteria following industry feedback on ambiguity.
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
