The Trump administration went to the G20 innovation summit this week determined to draw a bright line between American and European approaches to AI regulation. The EU's top tech official had a simple response: the gap is narrower than Washington likes to claim.

Different roads, similar destination

Henna Virkkunen, the European Commission's executive vice-president for tech sovereignty, security and democracy, told Axios at the summit in Chapel Hill, North Carolina, that the United States and Europe frequently end up addressing the same safety concerns about artificial intelligence. The difference is mechanism, not substance. Where Brussels legislates comprehensively through the EU AI Act, Washington relies on a patchwork of state laws, court cases and voluntary commitments.

"We see when it comes to technologies that in the USA the culture is different," Virkkunen said. "So often it's coming via court cases, but often very similar things anyway, very similar concerns." Her point cuts against the prevailing narrative that the US and EU have taken fundamentally divergent paths on AI. The substance of what regulators want, she argued, converges even if the process differs.

What the US is actually doing

The United States has not adopted sweeping federal AI legislation. That does not mean the federal government is absent from the field. The Trump administration has developed a voluntary framework for reviewing advanced AI models before release, including provisions that could give government officials access to models for up to 30 days. The White House does not plan to publish this framework. OpenAI's chief executive, Sam Altman, confirmed this week that the administration reviewed its Astra model under the scheme. "But we of course did it," he said, acknowledging the voluntary nature of the arrangement.

Alongside that framework, Washington shapes AI development through executive actions, national security restrictions and voluntary industry commitments. Individual states, meanwhile, are drafting their own rules with varying scope and ambition. Product changes are also being forced on companies through litigation. The cumulative effect is a regulatory environment that is less tidy than the EU's, but not necessarily less constraining.

Industry leaders push back

At the G20 summit, industry figures made their objections to the European approach plain. Michael Kratsios, director of the White House Office of Science and Technology Policy, asked a panel what foreign governments should and should not do on AI regulation. Elon Musk, chief executive of SpaceX, was blunt: "You have to have an environment that's relatively free of regulation, meaning that new things must be default legal as opposed to default illegal."

Musk's characterisation of the EU system, that things are generally default illegal, is a familiar Silicon Valley complaint. It reflects genuine frustration with the compliance burden of the EU AI Act and the Digital Services Act, under which ChatGPT now faces strict content moderation requirements. But it also understates the degree to which the US system imposes costs on AI developers through other channels, particularly litigation risk and the patchwork of state-level rules that companies must navigate.

The EU's own regulatory soul-searching

Virkkunen's remarks were not a full-throated defence of every line in the EU AI Act. She acknowledged that the bloc needs to simplify its rules and make Europe "easier, simpler, faster" for businesses. That language is a departure from the tone of her predecessor, Thierry Breton, who rarely missed a chance to champion regulatory ambition as a competitive advantage. It also reflects growing concern in Brussels that the accumulated weight of digital regulation, the AI Act, the Digital Services Act, the Digital Markets Act, the Data Act, is creating a compliance thicket that disadvantages European companies more than foreign rivals.

The admission matters because it comes from the person now responsible for implementing these rules. If the Commission's own tech chief is calling for simplification, the gap between European rhetoric and European reality may be wider than officials have previously conceded.

Pre-deployment testing: common ground

One area where US and EU approaches appear to be converging in practice is pre-deployment testing. Virkkunen noted that the EU is working on a framework for pre-deployment testing and trusted partner access, an area where the two sides face essentially the same policy questions. The Trump administration's voluntary review framework, under which OpenAI submitted its Astra model, addresses the same concern: ensuring that powerful models are assessed before they reach the public.

The difference is that the EU intends to codify these requirements in law, making them mandatory and enforceable. The US prefers voluntary commitments backed by the implicit threat of regulation if industry does not cooperate. Virkkunen's argument is that both paths lead to similar outcomes. The question for companies is whether they prefer the predictability of a legal framework or the flexibility, and uncertainty, of a voluntary one.

The practical question

Virkkunen framed the transatlantic difference as cultural rather than substantive: Americans trust courts and markets, Europeans trust legislation and institutions. There is truth in that, but it understates the commercial stakes. Companies operating in both markets face two distinct compliance regimes. A voluntary framework in Washington can be revised or withdrawn with a change of administration. The EU AI Act, for all its complexity, provides a stable legal baseline that companies can plan around.

The G7 and G20 discussions on pre-deployment testing and trusted partner access continue. The EU is developing its blueprint. The Trump administration is keeping its voluntary framework out of public view. Whether the two converge in practice, as Virkkunen suggests they already have, depends less on philosophy than on what actually happens when the next powerful model is ready for release.

People mentioned

Organisations

European Commission · White House Office of Science and Technology Policy · OpenAI · SpaceX