Is the EU’s AI Code of Practice under siege?

In the global race to shape the future of artificial intelligence (AI), the European Union has chosen not to run faster but to run wiser. Rather than competing for AI supremacy on technological grounds, Brussels is aiming to lead the world in something it knows best: setting rights-based regulatory standards. The EU’s Artificial Intelligence Act (AI Act), the world’s first comprehensive AI regulation, exemplifies this ambition. But recent developments suggest that this leadership role may be increasingly under threat, not only from within but also from across the Atlantic.

The latest controversy involves a voluntary Code of Practice on General Purpose AI (GPAI), currently being developed to complement the AI Act. Intended to help AI developers interpret and comply with the AI Act’s provisions, especially those regulating powerful, general-purpose models like ChatGPT or Gemini, the Code is envisioned as a crucial tool to bring legal clarity and establish responsible norms. However, the process is facing growing criticism on several fronts, with the US Mission to the EU even sending a formal letter urging the European Commission and national governments to reconsider the framework.

This move signals a new level of discomfort from the United States, which argues that the proposed code imposes excessive and impractical burdens on AI developers. US officials and industry leaders have warned that the voluntary framework, despite its non-binding nature, could go well beyond the scope of the AI Act and stifle innovation.

The European Commission had originally intended to publish the finalised code in early May, but missed the deadline, seemingly opting instead to extend consultations amid growing controversy. Civil society actors and investigative groups such as Corporate Europe Observatory and Lobby Control have criticised the drafting process, noting that big tech companies enjoyed structural advantages, while smaller stakeholders, such as rights holders, publishers and SMEs, were sidelined. Critics argue that the result could be a watered-down framework more aligned with corporate interests than public values. The code’s delay is thus not merely a bureaucratic hiccup, but a symptom of broader tensions over who gets to shape the rules of AI.

The US response reflects long-standing ideological differences over digital governance. Washington has historically been wary of the EU’s assertive regulatory approach, often viewing it as a veiled form of protectionism or a brake on innovation. Under the Trump administration, such criticisms became particularly vocal, with European digital regulations being likened to unfair taxation on American firms.

Despite the stance of the US administration, it has been reported that US firms are still engaging in the code’s development, signalling that while objections are real, compliance may be inevitable, especially for companies aiming to operate in the EU’s lucrative internal market. Interestingly, the US has also offered to send technical experts to support the EU in revising the draft, indicating a strategic interest in influencing the outcome rather than abandoning the process altogether.

Moreover, the intervention by the US comes at a delicate time. Critics have warned that the European Commission is showing a worrying tendency toward deregulation in the name of simplification and competitiveness, which is evident in recent legislative initiatives like the “omnibus proposal”. There are fears that the AI Code of Practice might become another casualty of this trend, leaving the door open to aggressive lobbying from tech giants keen to dilute accountability.

This drift risks undermining the very purpose of the AI Act, which is to bring legal certainty in upholding fundamental rights in the age of advanced AI while fostering innovation.The more fragmented and industry-driven the regulatory landscape becomes, the less effective the EU will be at achieving this goal.

At its core, the EU’s approach to AI is about defining global norms. Just as it did with the General Data Protection Regulation (GDPR), Europe aims to set a gold standard, which is grounded in transparency, accountability and human rights. But the transatlantic divide over digital policy seems to be widening, with different political and commercial priorities pulling in opposite directions.

The EU’s success now depends on maintaining the credibility and effectiveness of its own frameworks. If it allows the AI Code of Practice to be overly influenced or hollowed out, it may forfeit its position as the world’s moral compass in tech governance. The stakes are high, not just for Europe, but for global AI governance and the billions of people who will be affected by these technologies.

Whether companies ultimately endorse the Code remains uncertain. But what is clear is that the EU must stand firm in its commitment to rights-based regulation. With multilateral cooperation on shaky ground, Europe’s ability to act as a standard-setter is more crucial than ever. The AI Code of Practice, if designed inclusively and enforced meaningfully, can still be a cornerstone of trustworthy AI, serving not just European citizens but as a model for the world.