/s3/static.nrc.nl/wp-content/uploads/2026/05/07154005/070526BUI_2033538674_EUAI_Dragend.jpg "070526BUI_2033538674_EUAI_Dragend.jpg")
The lights in the European Parliament’s corridors rarely dim as quickly as the optimism surrounding the EU’s ambition to be the world’s ethical compass for artificial intelligence. After nine grueling hours of deliberation that stretched deep into Wednesday night, Brussels finally reached a consensus on the revised AI Act. But the agreement, born of exhaustion and high-stakes diplomacy, reveals a bruising struggle between the EU’s regulatory ideals and the cold reality of global competition.
What emerged was not a triumph of democratic deliberation, but a concession to power. The final hours of the session were defined by a series of behind-the-scenes maneuvers that critics are calling an “undemocratic turn.” At the center of the storm was a direct line between Berlin and the European Commission, highlighting a growing desperation within the bloc to prevent its industrial heartland from migrating toward the more permissive shores of the United States and China.
For years, the European Union has positioned itself as the “global regulator,” betting that the world would eventually adopt the “Brussels Effect”—where EU standards become the default global norm. However, as the pace of AI development accelerates, that bet is looking increasingly risky. The latest revisions to the AI Act suggest that Brussels is no longer leading the race; it is frantically trying to avoid being left behind.
The ‘German Pivot’ and the Industrial Loophole
The breakthrough on Wednesday night was less about compromise and more about a decisive push from Berlin. German Chancellor Friedrich Merz, facing dismal domestic approval ratings and a stagnating economy, successfully leveraged the Council to carve out significant exemptions for “industrial AI.” By arguing that machinery is already subject to existing sector-specific safety laws, Merz secured a victory that excludes a vast swath of industrial AI from the Act’s most stringent requirements.
The process of reaching this agreement was fraught. On April 29, a scheduled session was abruptly terminated after twelve hours by Commissioner Henna Virkkunen, following a directive from Björn Seibert, the chief of staff to Commission President Ursula von der Leyen. The move was seen as a direct translation of Merz’s demands into executive action.
Kim van Sparrentak, a Member of the European Parliament and shadow rapporteur, described the coordination between Merz and von der Leyen—both members of the German Christian Democrats—as a “one-two punch” that bypassed traditional democratic scrutiny. While van Sparrentak and her colleagues managed to claw back some safety guarantees for industrial AI at 4:00 a.m. On Wednesday, the overarching trend is clear: deregulation is winning.
Washington’s ‘Winning the Race’ Strategy
The pressure on Brussels is not merely internal. The United States, under a Trump administration focused on aggressive deregulation, has viewed the EU’s restrictive approach as a strategic opening. The U.S. AI action plan, launched in July 2025 under the explicit title “Winning the Race,” has functioned as both a carrot and a stick.
Jacob Helberg, U.S. Secretary of Economic Growth, has publicly characterized the European tech landscape as a “crisis,” a sentiment echoed by U.S. Ambassador to Brussels Andrew Puzder. Washington has not been subtle, pairing calls for deregulation with veiled threats of higher trade tariffs if European rules stifle American tech exports.
The impact of this pressure is most visible in the timeline. The implementation of rules for “high-risk AI”—systems involving facial recognition and critical infrastructure—which was originally slated for August 2 of this year, has been pushed back to December 2, 2027. This three-year delay provides a massive window for Big Tech to consolidate power before the regulatory hammer falls.
| Provision | Original Deadline | Revised Deadline | Impact |
|---|---|---|---|
| High-Risk AI Rules | August 2, 2025 | December 2, 2027 | Delayed oversight for critical infra/biometrics |
| Deepfake Labeling | – | December 2025 | Mandatory watermarking for AI content |
| ‘Nudifier’ Ban | – | Immediate/Phased | Criminalization of non-consensual AI nudity |
The Corporate Ultimatum: Guardrails vs. Rigidity
The political shift in Brussels is a direct response to a growing rebellion among Europe’s own industrial titans. In a recent open letter, seven of the continent’s most influential tech and industrial firms—including ASML, Airbus, Nokia, Mistral AI, and Siemens—warned that rigid requirements are stifling innovation. They argued that digital rules should serve as “flexible guardrails” rather than “rigid requirements.”
The threat is not theoretical. Roland Busch, CEO of Siemens, recently signaled at a trade fair in Hannover that the company would prioritize AI investments in the U.S. And China if the EU failed to soften its stance, mentioning a potential investment of one billion euros in the United States. For a bloc already struggling with an “innovation gap,” the prospect of its flagship companies fleeing is an existential threat.
This corporate leverage has left the European Commission in a state of hesitation. Insiders suggest the Commission has been waiting for a “global norm” to emerge, hoping to align its texts with Big Tech’s preferences rather than risking a standalone European standard that might alienate investors.
The Ethics Gap and the Fight Against ‘Nudifiers’
Despite the lean toward deregulation, the AI Act does retain some of its moral teeth, particularly regarding the protection of individuals. A significant victory in the latest text is the explicit ban on “nudifier” apps—AI systems that create non-consensual nude imagery. Marilena Raouna, Cyprus’s Vice Minister for European Affairs, noted that this move is critical for the protection of children and the fight against digital abuse.
The Act also targets the proliferation of sexually explicit deepfakes, a response in part to content produced by Elon Musk’s xAI chatbot, Grok. Starting this December, AI providers like OpenAI and Anthropic must watermark AI-generated images and videos to ensure transparency.
However, critics like Rebecca Petras of the AI accountability alliance Back Channel argue that these protections are a drop in the bucket compared to the lack of ethos in the “new generation” of tech entrepreneurs. While legacy firms like Microsoft may fund ethics boards, the current wave of AI pioneers operates with far fewer constraints, making the EU’s delayed regulations even more precarious.
As the AI Act moves toward formal ratification by the European Parliament and member states—a process now viewed as a mere formality—the Commission has another 18 months to write the actual technical standards. In a geopolitical climate where AI evolves by the week, the Brussels legislative machine is beginning to look less like a precision instrument and more like a legacy vehicle struggling to keep pace with a digital Ferrari.
The next critical checkpoint will be the formal ratification vote in the coming months, followed by the Commission’s first draft of the technical standards for industrial AI. These documents will determine whether the “guardrails” are truly flexible or if they have simply been removed.
Do you believe the EU is sacrificing its ethical standards for economic survival? Share your thoughts in the comments below or join the conversation on our social channels.
Disclaimer: This article discusses evolving legal frameworks and regulatory policies. It is provided for informational purposes and does not constitute legal advice.
