The European Union commenced enforcement this week of its landmark artificial intelligence regulation, the first comprehensive law aimed at governing AI technology. However, experts warn that the legislation, finalized in 2024, is already struggling to keep pace with rapid advances in AI capabilities, potentially limiting access to cutting-edge tools within the bloc.

The law sets a threshold defining an AI model as posing a "systemic risk" if it requires more than 10^25 mathematical operations during training, reflecting the technological landscape at the time of drafting. Since then, AI models have grown significantly more complex and powerful, raising questions about the law’s continued relevance. Observers note that certain provisions of the 144-page statute are vague, leaving companies uncertain about compliance and enforcement.

Key elements of the framework include prohibitions on social scoring and protections against exploiting vulnerable populations. The law also imposes rigorous safety requirements on AI systems used in critical areas such as employment decisions, loan approvals, and law enforcement. It mandates transparency measures, requiring firms to disclose when users are interacting with AI bots or viewing synthetic media.

Despite these aims, critics argue the law’s broad terminology creates ambiguity that complicates compliance. For instance, it bans manipulative or deceptive techniques that "materially" distort behavior and impair informed decision-making leading to "significant harm," but does not clearly specify what constitutes such harm or which techniques are covered. The classification of vulnerable groups is similarly expansive, encompassing not only age and disability but any individual in a “specific social or economic situation,” a phrase some say could apply to almost anyone.

The regulatory uncertainty disproportionately affects smaller companies and startups, which face high compliance costs. Meanwhile, larger technology firms allocate substantial legal resources to interpreting the law’s provisions. This environment has contributed to a slow AI ecosystem within Europe, with only a handful of frontier AI laboratories operating in the region, such as France’s Mistral. Their models reportedly lag behind those developed in the United States and China.

Several major tech companies have expressed reluctance to deploy their most advanced AI systems in the EU market. In 2024, Meta opted not to release its most capable models in Europe, citing unpredictable legal conditions. Apple, influenced by the Digital Markets Act and EU regulations, has restricted some AI features on iPhones sold in Europe. Google, too, is reportedly reconsidering its AI deployments on European devices.

While EU officials have touted the so-called "Brussels Effect," expecting global companies to align with European regulatory standards, some analysts suggest the opposite may be occurring. By imposing stringent and ambiguous rules, the EU risks limiting its own access to leading AI technologies and the economic opportunities they offer. This has raised concerns that the bloc’s regulatory ambitions may inadvertently stifle innovation and competitiveness in the evolving AI landscape.