The EU AI Act Has Teeth. The Question Is Whether It Bites.

The EU AI Act is moving from paper to practice. Prohibited AI systems are now banned. High-risk classification obligations are rolling in. The first real test of AI regulation at scale is underway — and it matters far beyond Europe.

The Act's risk-based framework sounds reasonable in a slide deck. In practice, the hard questions are just arriving. Who determines whether a system is "high-risk"? The deployer? The provider? A regulator staffed with a fraction of the technical expertise that built the system? The compliance burden falls heaviest on smaller European firms while US and Chinese giants have the legal teams and lobbying budgets to shape interpretation. Brussels has seen this before — GDPR's cookie banners became a running joke while actual data exploitation continued largely unchecked.

Still, the Act does things no other jurisdiction has attempted. Banning social scoring, emotion recognition in workplaces and schools, untargeted facial recognition scraping — these are lines drawn in law, not guidelines. The general-purpose AI provisions force some transparency about training data and compute, even if the specifics remain contested. Foundation model providers must disclose. That's new.

The enforcement question is everything. The European AI Office has a coordination mandate but limited direct power. National authorities vary wildly in capacity. Ireland's experience as a GDPR bottleneck for Big Tech — where enforcement was notoriously slow — looms as a cautionary template. Fines of up to 35 million euros or 7% of global turnover sound formidable. They mean nothing if they're never levied.

The redistribution angle: Without robust enforcement, regulation becomes a compliance cost for smaller players and a moat for incumbents. Well-resourced companies absorb the paperwork; startups and public-sector deployments struggle. The intended beneficiaries — people subject to AI decisions — only gain protection if regulators actually intervene. Rights on paper aren't rights in practice.

Will the EU fund enforcement proportional to the problem, or has it written a check its institutions can't cash?

#EUAIAct #AIRegulation #TechPolicy #Redistribution #AIGovernance