AI Act: Washington Challenges the Implementation of European Codes of Good Practice

A year after its official adoption, the European AI Act enters a decisive phase: its concrete implementation. But as the first codes of good practice are set to come into force by summer 2025, transatlantic tensions are intensifying.
Washington is ramping up diplomatic pressure to limit its scope, fearing overly strict regulations for its tech giants.
This standoff is no longer about the principle of AI regulation itself, but about the form it should take: between binding voluntary standards on the European side and pragmatic self-regulation on the American side, two visions are now clashing head-on.

Europe Implements the AI Act: Codes of Good Practice in the Spotlight

Adopted in March 2024, the AI Act made Europe the first power to establish a specific legal framework for regulating artificial intelligence.
The text is based on a graduated approach, proportional to the risk presented by each type of system: biometric surveillance, AI in recruitment, or in justice are among the most sensitive cases.

But for foundational models like GPT-4, Claude, Gemini, or their successors, Europe has opted for an intermediate mechanism: codes of good practice.
These texts, being finalized since early 2025, are meant to offer designers of large models:

• to precisely document their technical characteristics,
• to measure their environmental and societal impact,
• and to ensure transparency, bias management, and usage safety.

While not mandatory initially, the codes will be linked to a presumption of compliance, effectively making them a quasi-standard in relations between companies and European regulators.

For Brussels, this balanced approach allows for regulating usage without stifling innovation and establishing a culture of responsibility in the sector over abrupt bans, creating a trustful environment for businesses and citizens.

The American Response: Fear of a Global Regulatory Precedent

But on the American side, the implementation of these codes raises serious concerns.
Since February 2025, several representatives of the new Trump administration, staunchly pro-business and opposed to any excessive constraints on American tech giants, have expressed their strong opposition to what they see as disguised extraterritorial regulation.

Their arguments:

• The codes would impose an excessive regulatory burden on non-European actors, particularly Americans.
• They would create a commercial imbalance, hindering US companies' access to the European market.
• They risk setting a global normative precedent, compromising the freedom of innovation cherished by Washington.

Under Trump, the American doctrine has hardened, favoring a more sovereign and competitive approach. Europe is no longer seen as a regulatory partner but as a strategic rival in the global AI race.

The White House has reportedly used several diplomatic channels — including the Trade and Technology Council — to request relaxations in the final version of the codes, or even their requalification as non-prescriptive recommendations.

Brussels, for its part, stands firm. For the Commission, it is not about hindering the market but about setting minimal safeguards in a still largely unregulated sector.

The implementation of the AI Act, a year after its adoption, shows that the regulation of artificial intelligence has become a global strategic issue.
Far from mere technical divergences, the debate over the codes of good practice reveals a battle of influence between two digital visions: one of AI regulated from its inception, and one of AI free to grow before being regulated.

European companies — as well as subsidiaries of international players — must now anticipate compliance with upcoming obligations. Documentation, transparency, traceability: these are becoming implicit standards for operating in the European market.

This regulatory confrontation could well foreshadow a future global governance of AI, akin to what GDPR was for personal data. The question remains whether Europe can stay the course against pressure, and if its model will establish itself as a reference... or be bypassed in a global context of heightened competition.