EU Delays High-Risk AI Rules as US Lawmakers Propose Federal Framework

In June 2026, the global framework for artificial intelligence regulation underwent a structural realignment. While the European Union officially pumped the brakes on its landmark AI Act to give the industry more time, the United States saw a flurry of federal activity aimed at reining in a chaotic patchwork of state-level laws.[1][5]

The most concrete legislative shift occurred in Brussels. On June 16, the European Parliament approved the "Digital Omnibus on AI," a legislative package that fundamentally rewrites the compliance timeline for the world's first comprehensive AI law.[1][2]

The primary claim emerging from the Omnibus is a 16-month delay for core high-risk AI obligations. Under the approved text, the compliance deadline for stand-alone "high-risk" AI systems—categorized under Annex III, which includes employment screening tools and biometric identification—has been pushed from August 2, 2026, to December 2, 2027. Furthermore, obligations for AI embedded in regulated products, known as Annex I systems, are deferred to August 2028.[1][3]

However, legal analysts caution that this deferral is not a dismantling of the AI Act. The fundamental risk-based architecture remains intact, and the evidence shows that certain immediate prohibitions are strictly maintained to address acute public harms.[3]

Generative AI providers face immediate, unyielding deadlines for content moderation. The December 2, 2026, enforcement date remains firmly in place for Article 50 transparency obligations. This includes a strict mandate for machine-readable watermarking on AI-generated content and a total ban on AI systems that generate non-consensual intimate imagery, often termed "nudifiers," or child sexual abuse material.[2][3]

Across the Atlantic, the US regulatory approach is fracturing into three competing lanes: a deregulatory executive branch, an ambitious but stalled Congress, and highly active state legislatures.[4][5]

At the executive level, the White House has pivoted to a strictly voluntary oversight model for frontier AI. Following the January 2025 revocation of the Biden administration's mandatory AI reporting requirements, the Trump administration issued a new executive order on June 2, 2026, titled "Promoting Advanced Artificial Intelligence Innovation and Security."[5][6]

The June 2 directive establishes a voluntary 30-day pre-release review framework for "covered frontier models," allowing companies to submit models for federal cybersecurity testing. Crucially, the order explicitly prohibits federal agencies from imposing mandatory licensing, preclearance, or permitting requirements on AI developers, cementing a hands-off federal posture.[6]

In contrast to the executive branch, US lawmakers are attempting to nationalize AI governance to prevent state-by-state fragmentation. On June 4, Representatives Jay Obernolte and Lori Trahan released a 269-page bipartisan discussion draft titled the "Great American Artificial Intelligence Act of 2026."[4]

The draft legislation would impose transparency reports, critical safety incident reporting, and whistleblower protections on large-scale frontier developers. It also mandates a study on the "jawboning" of AI firms and censorship, reflecting bipartisan compromises aimed at balancing free speech with algorithmic safety.[4][5]

The most contentious provision of the Great American AI Act is its preemption clause. The bill would explicitly preempt state laws that specifically regulate the development of AI models for a three-year period, effectively freezing local legislative efforts to allow a federal standard to take root.[4][5]

This preemption targets a rapidly expanding web of state legislation. In 2026 alone, states considered over 1,500 AI-related bills. Colorado recently replaced its original AI Act with the broader Automated Decision-Making Technology Act, while Utah has enforced strict disclosure requirements for AI use in regulated professions.[5]

There is high uncertainty regarding the Great American AI Act's path to passage. Civil society organizations and labor unions have already mounted opposition to the preemption measures, arguing that federal gridlock should not invalidate state-level consumer protections. Without congressional action, the voluntary executive order and the state patchwork will remain the de facto US regulatory regime.[4][5]

Meanwhile, EU member states are actively building their enforcement infrastructure to prepare for the delayed deadlines. On June 17, Ireland—home to the EU headquarters of major tech firms like Google and Meta—published the Regulation of Artificial Intelligence Bill 2026.[7]

The Irish legislation establishes the independent "AI Office of Ireland" as the central coordinating authority and empowers the Competition and Consumer Protection Commission to levy administrative sanctions against non-compliant AI deployers. This signals that while the EU timeline has shifted, the bureaucratic machinery for enforcement is already coming online.[7]

For enterprise AI adopters and foundation model developers, the transatlantic divergence requires a bifurcated compliance strategy. In Europe, the legal requirements are clear but delayed, allowing companies to focus on the immediate December 2026 watermarking rules. In the US, companies must navigate voluntary federal guidelines while bracing for a potential collision between state enforcement and congressional preemption.[1][3][4][5]