EU AI Act delays high-risk rules but keeps disclosure deadlines

The European Parliament and the Council agreed on the Digital Omnibus on AI in May 2026, delaying selected EU AI Act requirements because technical standards and assessment routes were not ready. Standalone high-risk systems, including recruitment screening tools, now face a December 2027 compliance deadline. High-risk AI embedded in regulated products, such as medical devices and machinery, has until August 2028. National regulatory sandboxes move from August 2026 to August 2027.

The changes do not suspend the wider framework. General-purpose AI model obligations have applied since August 2025. Most transparency requirements, including informing people when they interact with AI and identifying generated content, retain their original deadline of 2nd August 2026.

Digital watermarking requirements move to 2nd December 2026 for content-generation tools already on the market before the original deadline. New tools launching after 2nd August 2026 must comply from launch. A new ban on systems generating illegal, exploitative imagery also takes effect on 2nd December 2026, primarily requiring providers to demonstrate safeguards against misuse.

Partners should inventory customers’ AI systems, distinguish standalone tools from embedded features, and identify high-risk uses requiring risk assessments, human oversight, data governance and documentation. UK-based businesses should also assess their exposure where they operate in the EU, serve EU citizens or deploy tools touching EU data. The extended timetable offers preparation time, not a blanket exemption from obligations already active or approaching.