Photo: Ron Lach Pexels
By Citizen of Europe | August 6, 2025
Section: Deep Dives & Analysis → Legal Corner
Europe just took another bold step in taming the machines. But not everyone’s playing along.
On August 2, the second enforcement phase of the EU Artificial Intelligence Act quietly took effect—targeting general-purpose AI models (GPAI) like those powering ChatGPT, Gemini, and open-source competitors. Companies must now adhere to strict transparency, safety, and data governance rules.
And while Google signed on, Meta refused.
This divide has thrown Europe into a high-stakes debate: is the EU securing the future—or regulating itself into irrelevance?
What’s this phase about?
Under the AI Act’s Article 52 & 53, companies deploying or developing general-purpose AI in Europe must now:
- Disclose training data sources
- Assess systemic risk to public safety or fundamental rights
- Implement safeguards against misinformation, hallucination, or bias
- Provide technical documentation for oversight
Those who don’t comply face fines of up to 7% of global turnover—a level previously reserved for GDPR violations.
To facilitate compliance, the Commission introduced a voluntary Code of Practice—not law, but considered a preemptive legal signal.
Google signed. Meta walked.
In late July, Google agreed to sign, committing to transparency mechanisms and reporting. But it warned that terms remain “broad and legally uncertain.”
Just days earlier, Meta declined outright, calling the code “vague, politically loaded, and functionally duplicative.”
Meta’s refusal places it at odds with the EU—and opens the door to regulatory scrutiny or fines down the line. Brussels sources say enforcement mechanisms are already being drafted for non-signatories.
Why the EU is doing this
The AI Act is the world’s first sweeping framework regulating artificial intelligence. It takes a risk-based approach, classifying AI systems by their impact on safety, rights, and democracy.
EU officials argue that opaque, unregulated models threaten everything from election integrity to human dignity.
“This isn’t about slowing down innovation—it’s about making sure innovation remains human-centered.”
— Margrethe Vestager, EVP for Digital Europe
They point to a flood of AI-generated misinformation, deepfakes, and black-box decision systems as proof that “move fast and break things” is no longer a responsible model.
The backlash: Will regulation sink innovation?
Critics warn that the Act—especially its ambiguity around “systemic risk” and “high-impact models”—could:
- Scare off AI startups
- Increase compliance costs for small and mid-size developers
- Push innovation toward the U.S., UAE, or offshore ecosystems
Even Bosch CEO Stefan Hartung—normally pro-EU—recently warned that “we are strangling homegrown AI while Silicon Valley accelerates.”
Meanwhile, the Trump administration in the U.S. has already scrapped most federal AI constraints, calling them “anti-business.” This raises fears that Europe’s principled approach may isolate it in global AI competition.
What’s really at stake?
This isn’t just about code—it’s about who writes the rules for the future.
- Can the EU become a global standard-setter for AI ethics and transparency?
- Or will its rules be seen as red tape, forcing developers to flee?
The next 6–12 months will be decisive. The Act’s third phase—covering biometric AI, surveillance tools, and military-use AI—will be even more politically explosive.
Final Word
Europe may be the first to try regulating frontier AI—but it won’t be the last. Whether it succeeds or stumbles will shape global AI norms for decades.
One thing is clear: we’re no longer just coding machines—we’re coding power.
Sources
- ITPro – What the August 2025 GPAI Rules Mean
- PC Gamer – Google Signs, Meta Refuses
- Carnegie Europe – The EU’s AI Power Play
- European AI Alliance – Implementation Updates
Disclaimer: This article is based on publicly available information and expert commentary as of August 6, 2025. It does not constitute legal advice or imply any unlawful conduct by companies or individuals mentioned. All parties are presumed compliant unless determined otherwise by competent authorities.
About the Author
