If you’ve built a compliance program around the EU AI Act’s August 2026 deadline for high-risk systems, the ground has shifted. Tech Policy Press reported on April 2 that positions adopted by the European Parliament and Council as part of the AI Omnibus would postpone core obligations to December 2027, or potentially as late as August 2028. August 2026 remains the current operative deadline until formally amended – but the direction of travel is now clear.
The AI Omnibus is the legislative vehicle. It’s a package of amendments to existing EU digital law, of which the AI Act high-risk provisions are one component. Michael McNamara, named as Co-Rapporteur for the Digital Omnibus on AI in the European Parliament, is among the key legislative actors shaping the positions now being debated. The Parliament’s proposals also include a ban on nudifier apps, an addition that sits alongside, not instead of, the delay provisions.
Two proposed new deadlines have emerged from the process: December 2027 and August 2028. Neither is final. These are legislative positions, not enacted dates. An organization that cancels its August 2026 compliance program today based on these proposals would be taking a real regulatory risk. The correct response is to track the legislative process while maintaining current program momentum.
The structural problem the delay creates is more significant than the timeline shift alone. The AI Act is not retroactive. A system placed on the market before the compliance deadline takes effect is not required to comply after the fact. Analysts warn that this means AI systems deployed before whichever new deadline ultimately takes effect could permanently remain outside the law’s oversight requirements, not temporarily exempt, but structurally unreachable. That isn’t a bug in the law’s design; retroactivity would create its own problems. But combined with a two-year delay, it creates a window during which high-risk systems can enter the EU market without meeting the requirements the law was written to impose.
The debate also surfaces questions about which systems count as high-risk in the first place. According to Tech Policy Press reporting, there’s active discussion about overlap between the AI Act and sector-specific legislation. Medical devices, toys, and connected cars are among the categories that may be handled under separate sectoral rules rather than the AI Act’s high-risk framework. The scope of what the delay actually applies to is still being defined.
What to watch: whether the Parliament and Council reconcile their positions on the December 2027 versus August 2028 options, and whether the AI Omnibus moves to a final vote. The Tech Policy Press analysis notes that critics argue the delay risks weakening the law’s impact at a critical moment for AI governance. That framing reflects the analytical community’s concern, not a settled factual finding, but it maps the political terrain the legislation is crossing.
The compliance planning takeaway is precise: August 2026 is still the operative deadline. Don’t stand down. Do begin modeling what a December 2027 or August 2028 transition looks like for your high-risk system inventory, including which systems would be deployed before any new deadline and therefore outside the law’s reach entirely. That inventory question isn’t hypothetical anymore.