CHAPEL HILL, North Carolina — G20 countries have agreed to a set of non-binding principles calling for a lighter approach to regulating artificial intelligence and other emerging technologies, giving the Trump administration support for its campaign against what it sees as excessive technology rules.
The agreement was reached on Wednesday at a two-day G20 Innovation Ministerial in Chapel Hill, North Carolina, hosted by the US Commerce Department and the White House Office of Science and Technology Policy. US Commerce Secretary Howard Lutnick announced the consensus at the close of the meeting.
Known as the Carolina Principles, the framework calls on governments to take a sector-by-sector approach to new technology rules rather than create broad new regulatory systems specifically for AI. It also encourages closer cooperation between governments and technology companies when assessing emerging technologies.
The principles are not legally binding. They are expected to be put before G20 leaders for formal adoption at their December summit in Florida.
Reaching agreement among the G20 was not straightforward, Lutnick said. “When you bring global leaders together there’s always lots of politics,” he said, describing the negotiations as a substantial effort.
The agreement represents a diplomatic win for Washington at a time when governments are taking very different approaches to AI. The European Union has adopted tougher rules in several areas, while the Trump administration has argued that regulation should not slow the development or deployment of new technology.
Michael Kratsios, the White House science and technology adviser, said the US wanted countries to focus on economic growth and innovation rather than build new regulatory structures around every new technology.
The US position also has strong support from the American technology industry. Executives from companies including Nvidia, OpenAI, SpaceX and others took part in the meeting, making the gathering unusual for the extent of direct industry involvement in discussions among governments.
Nvidia chief executive Jensen Huang argued that governments should regulate actual harm caused by AI rather than risks that remain hypothetical. “The technology is in its formative stages,” Huang said, adding that companies themselves should take responsibility for developing their systems safely.
The push for fewer restrictions comes as AI systems are becoming more capable and governments face pressure to address risks ranging from cybersecurity to misinformation and the use of autonomous systems.
US officials have argued that imposing broad rules too early could put American companies at a disadvantage and slow the adoption of technologies that could raise productivity. Reuters reported earlier this week that the administration was urging G20 members to reserve new regulation for novel problems rather than treating each new AI development as a separate regulatory challenge.
The agreement received backing from China and Russia as well as US allies, despite wider disagreements over technology and trade. Washington and Beijing are competing for an advantage in AI, with Chinese companies making rapid progress in developing increasingly capable models.
That rivalry was evident at the meeting. Lutnick said US and Chinese officials had held discussions on ways to cooperate, but he gave no indication that Washington was prepared to loosen its restrictions on China’s access to advanced semiconductors and chip-making equipment.
The US has spent years restricting China’s access to the most advanced chips used to develop AI systems. Those controls remain separate from the G20 framework, which deals primarily with regulation and technology policy rather than US export restrictions.
Relations with Europe have also been strained over AI policy. The Trump administration has pressed the European Union to ease digital regulations that American technology companies oppose, using tariffs and other measures as leverage.
Henna Virkkunen, the European Commission’s executive vice president for technology sovereignty, said Europe wanted to encourage innovation but also needed governments to work together on the risks created by increasingly capable AI systems.
Recent incidents involving AI systems operating with limited human supervision have added to those concerns. Virkkunen said governments needed to cooperate on how such risks should be handled rather than leaving the issue entirely to individual companies.
The Carolina Principles therefore leave considerable room for national governments to determine how AI should be regulated. Their immediate significance lies less in creating a common rulebook than in establishing a shared preference among the G20 for regulation that does not automatically expand as the technology develops.
