EU AI Act Faces Compliance Hurdles and Mounting Pressure for Delay

If you've tuned in over the past few days, the European Union’s Artificial Intelligence Act—yes, the much-debated EU AI Act—is once again at the center of Europe’s tech spotlight. The clock is ticking: obligations for providers of general purpose AI models entered into force on August 2nd, and by next summer a whole new layer of compliance scrutiny will hit high-risk AI. Yet, as Politico and Pinsent Masons have confirmed, several member states, Germany included, are lagging on the practical steps needed for effective implementation, thanks in part to political interruptions like Germany’s unscheduled elections and, more broadly, mountains of lobbying from industry giants worried they might lose ground to the U.S. and China. So, what’s truly new about the EU AI Act, and where does it stand today? First, let’s talk risk. The Act carves AI into four risk buckets—unacceptable risks like social scoring are banned outright. High-risk AI, think of systems in healthcare, finance, hiring, or biometric identification, are required to jump through regulatory hoops: they need high-quality, unbiased data, thorough documentation, transparency notices, and human oversight at pivotal decision points. Fines for non-compliance can be up to €35 million, or a hefty 7% of global revenue. The teeth are sharp even if enforcement wobbles. But here’s the present tension: there’s mounting pressure for a delay or “grace period”—some proposals floating around the Council hint at a pause of six to twelve months on high-risk AI enforcement, seemingly to give businesses breathing room. Mario Draghi criticized the law as a “source of uncertainty,” and Henna Virkkunen, the EU’s digital chief, is pushing back hard against delays, insisting that standards must be ready and that member states should step up their national frameworks. Meanwhile, the European Commission is busy publishing Codes of Practice and guidance for providers—like the voluntary GPAI Code released in July—that promise reduced administrative burdens and a bit more legal clarity. There’s also the AI Office, now supporting its own Service Desk, poised to help businesses decode which obligations actually bite and how to comply. The AI Act doesn’t just live in Brussels; every EU country must set up its own enforcement channels, with Germany giving more power to regulators like BNetzA, tasked with market surveillance and even boosting innovation through AI labs. Civil society groups like European Digital Rights and AccessNow are demanding that governments move faster to assign competent authorities and actually enforce the rules—today, most member states haven’t met even the basic deadline. At the innovation end, Europe’s AI Continent Action Plan is trying to spark development and scale up infrastructure with things like AI gigafactories for supercomputing and data access—all while ensuring that SMEs and startups aren’t crushed by compliance bureaucracy. So listeners, in this high-tension moment, Europe finds it This content was created in partnership and with the help of Artificial Intelligence AI.