Inside the G20 Push to Keep AI Lightly Regulated

Inside a North Carolina conference hall on Wednesday, trade and commerce ministers from the world's largest economies heard a coordinated pitch from the people building the technology: adopt AI fast, let companies police themselves, and make it easier to build the data centers that train and run it.

The forum was the G20 Innovation Ministerial, hosted by US Commerce Secretary Howard Lutnick in a state now positioning itself as the center of gravity for AI infrastructure. The roster of speakers was unusually heavy: OpenAI's Sam Altman, Nvidia's Jensen Huang, and Anthropic executives on stage for back-to-back fireside chats, with Elon Musk reportedly joining virtually from the University of North Carolina.

The executives landed in the room the same week the European Union did the opposite. On Monday, two days before the North Carolina meeting, the EU reclassified ChatGPT as a "very large online search engine" under platform rules, opening a stricter review track on OpenAI specifically. The simultaneity is the point: industry is pressing G20 ministers to install the lightest version of AI rules as the global default just as Brussels is moving the other way.

The substantive ask was governance itself: how AI should be ruled globally. Altman told ministers that adopting the technology was "non-negotiable" and returned repeatedly to an analogy. "We want to put sufficient constraints on our system to avoid catastrophic risks, but we really do wanna respect all of your sovereignty and that you're going to use it in very different ways, and we personally may not always like it, but I don't like everything that people do with electricity either," he said, per Gizmodo's on-site reporting.

That is the same light-touch posture the Trump administration has been pushing in pre-meeting briefings: a general-purpose utility whose governance should sit with national governments and the companies building it. It is also the position Brussels is moving against this week.

Jensen Huang, who also pressed regulators to loosen the rules around AI infrastructure, was more direct about the stakes. "Every single country should advocate for AI diffusion or the adoption of AI into your industries," he said, per Gizmodo. "The worst outcome is that you don't take advantage of it, that you are left behind. That is the single worst outcome. That outcome can happen if we talk about the technology with fear, with uncertainty."

His answer to that fear was to push it back onto industry. Huang warned officials against regulating "hypothetical, theoretical harm" and said companies should "take it upon themselves to develop the technology safely." That phrasing, companies as the safety regulator of last resort, is the procedural lever the executives are trying to write into a G20-shaped communiqué.

For Altman, the economic case pointed to the same destination. He framed AI's upside as "too high to ignore" and predicted "the greatest boom in entrepreneurship and the creation and development of small businesses the world has ever seen" if ministers moved quickly, per Gizmodo.

The harder subtext is the data-center buildout. Nvidia and the cloud providers now sit at the center of an investment cycle that depends on cheap power, fast permitting, and a permissive rules environment. Asking ministers to ease the rules around new construction is a direct ask from a company that sells the chips those data centers will run on. The lobbying is for the next capex cycle, not the next election.

The EU's enforcement track, by contrast, keeps moving on its own clock. If the bloc's "very large online search engine" reclassification survives a procedural challenge and the resulting obligations bite, Brussels will have set a template other regulators can import by reference. The G20 communiqué is non-binding. Its value, if it lands, is procedural: a US-styled "light-touch utility" positioning written into a room that also includes the EU's principal regulatory competitors.

The text from the chair's office is the procedural artifact. If the communique treats AI as a general-purpose technology whose risks belong to companies and national governments, the executives have roughly a year of cover. If it is vague, the EU track keeps running on its own clock, and the North Carolina meeting goes into the record as a speaking tour the industry could afford but could not move.