The European Union has begun exercising sweeping new enforcement powers under its landmark Artificial Intelligence Act, giving Brussels the authority to inspect AI models before they are released in the bloc, restrict market access and levy substantial fines on providers who fall short of the rules, in a move that cements the Commission’s position as the world’s most assertive AI regulator.
Under the powers, which took effect earlier this month, the European Commission can now demand the right to evaluate AI models before they are made publicly available across the EU, a level of pre-market scrutiny that goes considerably further than anything currently in place in the United States or most of Asia. Providers found in breach of the rules face fines of up to €15 million or 3 percent of their annual global turnover, whichever figure is higher, giving the regulation real financial teeth against even the largest technology companies.
The expanded authority has placed some of the world’s most prominent AI developers, including Anthropic, OpenAI and Google, squarely in Brussels’ line of sight. Reports indicate the EU has opened discussions with both Anthropic and OpenAI following a series of cyberattacks linked to their models, with OpenAI confirming it is in contact with the bloc’s dedicated AI Office to address the concerns. Google, meanwhile, has already felt the weight of European regulators this year, having been hit with a €1 billion fine in July after officials found the company had given preferential treatment to its own services over competitors, a decision that prompted US President Donald Trump to threaten the EU with a “substantial” retaliatory tariff.
The rollout of these powers marks the latest stage in a carefully phased implementation of the AI Act, which was first passed in 2024 as the world’s first comprehensive legal framework specifically designed to govern artificial intelligence. Rules banning certain high-risk AI practices and establishing baseline AI literacy requirements came into force in February 2025, followed by obligations for general-purpose AI models later that year. Transparency requirements, which help users understand when they are interacting with AI systems or AI-generated content, took effect at the start of this month, giving citizens across the bloc clearer visibility into where and how the technology touches their daily lives.
Not all elements of the Act have moved forward on schedule. The Commission has pushed back some of the law’s more demanding high-risk obligations to December 2027, a delay officials have framed as a technical adjustment rather than a retreat from regulatory ambition. Executive Vice President Henna Virkkunen has argued the extra time will allow companies and regulators alike to develop clearer guidance and technical standards before the toughest requirements take hold, a decision that also reflects the Commission’s growing sensitivity to concerns that overly rigid regulation could hamper Europe’s competitiveness in the global AI race.
That competitiveness concern has become a recurring theme in Brussels’ approach this year. In a parallel move, the Council presidency and European Parliament negotiators reached a provisional agreement in May to streamline parts of the AI Act as part of a broader package aimed at reducing administrative burdens on European companies, while preserving the law’s core protections. Officials have framed the changes as reinforcing rather than undermining the EU’s digital sovereignty, arguing that a smoother, more harmonised set of rules will ultimately make the bloc a more attractive place for AI innovation, not a less regulated one.
The stakes extend well beyond Europe’s borders. Because many global AI providers build their products to a single, unified standard, disclosure requirements introduced under the EU’s rules are expected to ripple outward, shaping how AI companies operate even in markets outside the Union. At the same time, the law’s strongest protections, including the new inspection and enforcement powers, remain confined to products and services offered within the EU itself, leaving a patchwork of regulatory standards across different regions.
For companies like Anthropic, OpenAI and Google, the message from Brussels is unambiguous: operating in the world’s largest single market now comes with a level of scrutiny that few other jurisdictions currently attempt to match, and one that is likely to intensify as the AI Office builds out its capacity over the coming years. NEXT ARTICLE
