We're living through a peculiar moment in AI regulation. The European Union's Artificial Intelligence Act just came into force this past August, and already the European Commission is frantically rewriting the rulebook. Last month, on November nineteenth, they published what's called the Digital Omnibus, a sweeping proposal that essentially admits the original timeline was impossibly ambitious.
Here's what's actually happening beneath the surface. The EU AI Act was supposed to roll out in phases, with high-risk AI systems becoming fully compliant by August twenty twenty-six. But here's the catch: the technical standards that companies actually need to comply aren't ready. Not even close. The harmonized standards were supposed to be finished by April twenty twenty-five. We're now in December twenty twenty-five, and most of them won't exist until mid-twenty twenty-six at the earliest. It's a stunning disconnect between regulatory ambition and technical reality.
So the European Commission did something clever. They're shifting from fixed deadlines to what we might call conditional compliance. Instead of saying you must comply by August twenty twenty-six, they're now saying you must comply six months after we confirm the standards exist. That's fundamentally different. The backstop dates are now December twenty twenty-seven for certain high-risk applications like employment screening and emotion recognition, and August twenty twenty-eight for systems embedded in regulated products like medical devices. Those are the ultimate cutoffs, the furthest you can push before the rules bite.
This matters enormously because it's revealing how the EU actually regulates technology. They're not writing rules for a world that exists; they're writing rules for a world they hope will exist. The problem is that institutional infrastructure is still being built. Many EU member states haven't even designated their national authorities yet. Accreditation processes for the bodies that will verify compliance have barely started. The European Commission's oversight mechanisms are still embryonic.
What's particularly thought-provoking is that this entire revision happened because generative AI systems like ChatGPT emerged and didn't fit the original framework. The Act was designed for traditional high-risk systems, but suddenly you had these general-purpose foundation models that could be used in countless ways. The Commission had to step back and reconsider everything. They're now giving European regulatory sandboxes to small and medium enterprises so they can test systems in real conditions with regulatory guidance. They're also simplifying the landscape by deleting registration requirements for non-high-risk systems and allowing broader real-world testing.
The intellectual exercise here is worth considering: Can you regulate a technology moving at AI's velocity using traditional legislative processes? The EU is essentially admitting no, and building flexibility into the law itself. Whether that's a feature or a bug remains to be seen.
Thanks for tuning in to this week's deep dive on European artificial intelligence policy. Make sure to subscribe for more analysis on how regulation is actually shaping the technology we use every day. This has been a quiet please production, for more check out quiet please dot ai.
Some great Deals
https://amzn.to/49SJ3QsFor more check out
http://www.quietplease.aiThis content was created in partnership and with the help of Artificial Intelligence AI
