If you're building a product with AI and you've been quietly hoping someone would give you more time before the regulations kick in, good news. The EU just did exactly that.
Last week, the European Commission announced that the AI Act's deadlines for high-risk AI systems are being pushed back. What was supposed to happen this August now moves to December 2027, with some categories extending to August 2028.
That's 18 months of breathing room that didn't exist a week ago.
But before you move on, it's worth understanding what this actually changes, and what it doesn't.
What changed and what didn't
The AI Act was always going to apply to anyone building AI products that touch certain areas: employment tools, education software, biometrics, critical infrastructure. That hasn't changed. What changed is when you need to be ready.
The compliance deadline for most high-risk systems moved from August 2026 to December 2027. If your product falls into categories that intersect with existing EU safety rules, you get even more time, until August 2028.
The reason for the delay is straightforward. European businesses pushed back hard, arguing that the original timeline was putting them at a disadvantage against US and Chinese competitors. The Commission agreed, and wrapped it into what they're calling the AI Omnibus package, with simplified rules, delayed timelines, and reduced bureaucracy.
What doesn't move: the enforcement infrastructure. From August 2026, EU member states will have their market surveillance authorities operating. The penalties framework goes live. In Poland, the government is currently building KRiBSI, a national AI oversight body with the power to pull a dangerous AI product from the market within seven days. The regulators are getting ready even if the deadlines moved.
The trap in the delay
Here's the thing about 18 months of extra time: it disappears fast when you're building something.
The AI Act's requirements around transparency, human oversight, and documentation aren't things you add at the end of a project. They're architectural decisions. Who can challenge a decision your AI makes? How is the model monitored after launch? What happens when it produces a wrong output? These questions need answers before the first line of code, not before the first audit.
Think about something as simple as an AI feature that scores job applicants or flags customer accounts for review. Under the AI Act, users will have the right to request a human review of that decision. If your system wasn't built to support that, if there's no audit log, no explainability layer, no way to trace why the AI made a particular call, retrofitting it later is expensive and slow. It's the kind of thing that takes weeks to add and months to test properly.
The companies that will struggle most in 2027 are the ones that interpreted this delay as permission to ignore the problem. The ones that will be ready are the ones that use this window to build the right foundation.
What we think about it
Piotr Czyż, CEO of Rocksoft and a software engineer with over 15 years of experience, sees the delay as a real opportunity for Europe:
"Deferring the AI Act is a smart move to embrace AI innovation in Europe. Although we are behind the US or China, it doesn't mean we're less smart than they are, but politics and regulations kill European startups. We hope it will spark many interesting initiatives that were outsourced across the ocean. Looking forward to the coming months."
We share that optimism. European founders are capable of building world-class AI products. What they've often lacked isn't talent, it's the runway to move without regulatory uncertainty hanging over every decision.
That runway now exists. The question is what you do with it.
The AI Act delay doesn't change the fact that nearly half of AI applications contain critical security gaps. It doesn't change the fact that 85% of AI initiatives never make it to production. It doesn't change the fact that an app that works in a demo can expose user data when real people start using it.
What it does is give you time to build correctly. And that's exactly what our AI powered product development service is for, helping founders build AI products that are fast to ship, secure in production, and architecturally ready for whatever regulatory environment comes next.
The short version
The EU AI Act has been delayed, not cancelled. You have until December 2027 for most high-risk systems. Enforcement infrastructure goes live this August. The regulators are getting ready.
Use the extra time well. The founders who will be ready in 2027 are building the right foundations today.
Building an AI product and not sure where you stand? Let's talk.
