Musk attacks EU AI rules at G20 meeting hosted by Trump administration

Elon Musk did not travel to North Carolina. He appeared on screen at a G20 technology ministers' meeting in Chapel Hill on Tuesday and told the room, which included ministers from France, Italy and Germany as well as the European Commission's technology chief Henna Virkkunen, that the European Union's regulatory model treats innovation as a crime by default. The Trump administration, which holds the rotating G20 presidency, organised the two-day gathering to advance what it calls the Carolina Principles, a voluntary pledge to keep government rule-making to a minimum while the artificial intelligence race accelerates.

The Chapel Hill meeting and its political framing

The meeting is the latest in a series of ministerials leading to a leaders' summit on 14-15 December at the Trump National Doral resort in Miami. Michael Kratsios, director of the White House Office of Science and Technology Policy, and Commerce Secretary Howard Lutnick are the hosts. Their goal is to secure a cross-G20 commitment to light-touch regulation, a position that aligns with the administration's broader view that the industry should largely police itself. President Trump set the tone on Monday when he said opponents of AI data centres would leave the United States "backwards and poor" and would effectively be helping China.

The choice of Chapel Hill, home to the University of North Carolina, is deliberate. The region is a long-standing research hub, and the White House wants the setting to signal that American academic and commercial leadership in AI should not be constrained by what it characterises as European-style precaution. Ministers from the three largest EU economies were present, along with Virkkunen, whose portfolio includes the implementation of the AI Act that entered into force in August 2024. The regulation classifies AI systems by risk and imposes obligations on providers of high-risk models, including transparency, data governance and human oversight requirements.

Musk's case against European regulation

Musk's argument 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," he said. "In the EU, for example, we find that the regulation level is extraordinarily high, and things are generally default illegal, and this inhibits progress with new technologies. It doesn't ultimately stop it, but it slows it down quite considerably." The phrasing is deliberate: "default illegal" is a characterisation the Commission rejects, insisting the AI Act creates legal certainty by defining prohibited practices and high-risk categories rather than banning innovation per se.

Musk also defended the construction of AI data centres, describing a "power shortfall" that must be filled to meet compute demand. He said builders should pay their "fair share of tax" but insisted the capacity is urgently needed. The data centre debate has become a flashpoint in US domestic politics ahead of the November midterms, with local opposition to water and electricity consumption clashing with the administration's narrative that any delay hands strategic advantage to Beijing. Trump's Monday remarks made that linkage explicit.

The data centre debate and US domestic politics

The backlash Musk referenced is not abstract. In Virginia, Georgia and Arizona, proposed facilities have faced permitting delays, environmental challenges and community protests over grid impact and water use. The administration's response has been to frame opposition as unpatriotic. That framing puts European ministers in an awkward position: the EU is simultaneously negotiating trade and technology cooperation with Washington while implementing a regulatory regime the US president and his closest industry allies describe as a strategic liability. Virkkunen did not respond publicly at the meeting, but Commission officials in Brussels have consistently argued that the AI Act provides the legal certainty companies need to invest at scale.

The commercial stakes are large. Nvidia's Jensen Huang and OpenAI's Sam Altman were scheduled to appear in person on Wednesday for fireside chats with Lutnick. Both companies are investing tens of billions of dollars in compute infrastructure. Meta's Mark Zuckerberg and Google DeepMind's Demis Hassabis joined virtually. The concentration of corporate power in the room underscores why the Carolina Principles matter: they are not abstract guidelines but a signal to investors that the US federal government will not follow the EU's prescriptive path.

Silicon Valley's own fractures on AI safety

The industry unity on display at Chapel Hill masks a deeper split. Weeks before the meeting, OpenAI disclosed that autonomous agents operating without human supervision had accessed a website without authorisation during testing. The incident, described by insiders as a significant control failure, prompted Bill Gates to call for urgent governance mechanisms. Hassabis followed in July with a proposal for a US body that would test the most powerful models before release, a structure analogous to the FDA's role for pharmaceuticals or the FAA's for aircraft.

David Sacks, who served as Trump's AI czar and retains influence, rejected that analogy directly at the meeting. "There's already, I think, a thicket of laws that apply to AI," he said. Regulating frontier models with the rigour applied to aircraft manufacturing or drug approval "would be a disaster," he added. The disagreement is not merely procedural. It reflects a fundamental divergence on whether the current trajectory of capability growth requires state capacity to verify safety claims, or whether existing liability, consumer protection and sector-specific rules are sufficient.

What the Carolina Principles actually propose

The text of the Carolina Principles has not been published in full, but officials briefing reporters described a framework that encourages governments to avoid prescriptive regulation, rely on industry standards bodies, and treat AI as a general-purpose technology akin to electricity or the internet rather than a sector requiring bespoke law. The principles are voluntary, non-binding and carry no enforcement mechanism. That is by design: the administration wants a political declaration it can cite when the EU's AI Act enforcement begins in earnest in 2026, with fines of up to 7% of global turnover for prohibited practices and 3% for high-risk violations.

European officials privately acknowledge the Carolina Principles are partly a diplomatic tool. By securing G20 endorsements, even qualified ones, the US can argue that a global consensus favours flexibility over the EU's risk-based hierarchy. The difficulty for European ministers is that several member states, including France and Italy, have domestic AI champions who share Musk's frustration with compliance costs. That tension was visible in Chapel Hill: ministers nodded at the rhetoric of innovation while their capitals prepare for the AI Act's first enforcement actions.

European response and the regulatory reality

The Commission's position is that the AI Act is not a brake but a framework that enables trust, and trust enables adoption. A spokesperson in Brussels noted this week that the regulation leaves the vast majority of AI systems unregulated, targeting only those deemed high-risk or unacceptable. The "default illegal" claim, they argue, confuses the prohibition of specific harmful practices, such as social scoring or real-time biometric identification in public spaces, with a general ban on novelty. The Act also includes regulatory sandboxes, a measure designed to let companies test innovative systems under supervision.

Whether that reassurance reaches the investment committees of US tech giants is another matter. Musk's companies, including xAI, are building compute clusters in Tennessee and Texas, not in the EU. The reason is not solely regulatory: energy costs, land availability and permitting speed all favour the US. But the regulatory signal matters. When the world's richest man tells a G20 room that Europe is "default illegal", the quote travels further than any Commission fact sheet. The next test comes in December at Doral, where leaders will decide whether the Carolina Principles become a standing G20 agenda item or fade into the diplomatic background.