US pushes light-touch AI rules at G20 as tech chiefs demand more data centers

The Chapel Hill G20 meeting crystallizes a growing split between the US and Europe over how hard to lean on AI regulation. US officials are pushing the “Carolina Principles” which explicitly argue against building bespoke AI regulatory regimes and instead favor tweaking existing rules. That posture aligns cleanly with the interests of OpenAI, Anthropic, Meta and others who want flexibility to ship new frontier models without going through a fresh licensing gauntlet.

At the same time, CEOs like Mark Zuckerberg and Elon Musk used the forum to campaign for massive expansion of AI data‑center capacity and to attack EU‑style oversight as a drag on innovation. Musk is already talking about an impending energy crunch created by AI loads; Zuckerberg is arguing that opposition to data‑center siting is economically irrational. The subtext is clear: keep the regulatory lane open and let capital and hardware scale as fast as possible.

For AGI timelines, a coordinated G20 push toward light‑touch rules would tilt the world toward faster deployment, even if individual jurisdictions like the EU keep higher bars. It also signals that, in the US at least, the political center of gravity is still closer to industry than to risk‑averse safety advocates. That makes voluntary frameworks like OpenAI’s Preparedness thresholds even more important, because formal law may lag the capabilities frontier by years.