When the AI Act was adopted, Europe positioned itself as the first major jurisdiction attempting comprehensive horizontal regulation of artificial intelligence. The legislation was widely presented as the global benchmark for “trustworthy AI”: risk-based, rights-oriented, and capable of setting international standards beyond the EU itself.
Less than two years later, however, the EU is already modifying parts of its implementation framework.
The recently agreed “Digital Omnibus on AI” delays certain obligations for high-risk systems, simplifies requirements for smaller companies, clarifies overlaps with sectoral legislation, and expands access to regulatory sandboxes. The political message from Brussels is that simplification does not weaken the AI Act itself. Yet the speed at which implementation adjustments became necessary is revealing in its own right.
The issue is not that the AI Act’s core principles suddenly became irrelevant. Most large companies operating in Europe already accept that some level of AI governance is inevitable, particularly in areas involving biometric systems, critical infrastructure, employment decisions, healthcare, financial services, or public-sector applications.
The problem is operational reality.
The AI landscape evolved dramatically between the moment policymakers first drafted the legislation and the moment companies actually began preparing for implementation. Generative AI systems expanded at extraordinary speed, foundation models rapidly reshaped commercial deployment patterns, and businesses suddenly found themselves attempting to interpret compliance obligations for technologies that barely existed when the original negotiations began.
This created an uncomfortable mismatch between legislative timelines and technological acceleration.
Many companies — particularly SMEs and mid-sized firms — struggled not necessarily with the principle of AI governance, but with the cumulative complexity emerging around documentation, risk classification, transparency obligations, human oversight requirements, cybersecurity expectations, and interactions with existing sector-specific regulation.
European Commission – AI regulatory sandboxes
The Commission’s simplification package implicitly acknowledges something European regulators rarely state openly: implementation capacity has become a strategic issue in itself.
This is not unique to AI. Similar tensions have emerged across sustainability reporting, due diligence, ESG disclosures, cybersecurity legislation, and supply-chain regulation. In several areas, the EU is increasingly discovering that highly ambitious frameworks can generate serious operational friction once companies begin translating broad legal principles into internal systems, governance structures, supplier management, technical controls, and audit processes.
AI simply magnifies the problem because the technology itself evolves faster than most regulatory cycles can realistically accommodate.
The irony is that the AI Act was originally intended partly as a competitiveness tool. European policymakers hoped that regulatory clarity would create trust, encourage adoption, and position Europe as the leading jurisdiction for safe and reliable AI deployment. Instead, parts of the debate have increasingly shifted toward concerns that Europe risks building the world’s most sophisticated compliance framework around technologies largely developed elsewhere.
That concern is becoming harder to ignore.
The United States continues to dominate commercial AI scaling. China combines state-backed industrial coordination with massive domestic deployment capacity. Europe, by contrast, often appears strongest in governance architecture rather than technological scale itself. The more complex implementation becomes, the greater the risk that compliance capacity — rather than innovation capacity — becomes Europe’s primary competitive differentiator.
At the same time, the political pressure to simplify should not be misunderstood as deregulation.
The Omnibus package simultaneously introduces stricter prohibitions around abusive AI applications such as non-consensual sexually explicit deepfakes and so-called “nudification” systems.
This illustrates the increasingly selective direction of European digital governance: less focus on blanket administrative expansion, more focus on politically sensitive and high-risk uses where public harm is easiest to demonstrate.
What is changing is therefore not the EU’s willingness to regulate AI, but its growing recognition that regulatory credibility depends heavily on whether implementation remains economically and operationally workable.
That may ultimately become one of the defining policy challenges of the next decade. Europe is entering an era in which competitiveness, simplification, industrial policy, and regulatory ambition are no longer separate debates. Increasingly, they are colliding inside the same legislative frameworks.
The AI Omnibus may therefore matter less for the individual delays and procedural adjustments it introduces than for what it quietly signals: even Europe’s flagship digital regulation now has to adapt to the reality that governing fast-moving technologies is not only about writing rules, but about whether entire economic systems are realistically capable of absorbing them.