22 countries call for keeping AI "under human control" and propose creating a global oversight body

A group of 22 countries adopted a declaration on Monday, on the sidelines of the UN General Assembly in New York, stating that artificial intelligence "must remain under human direction, oversight and control." The text acknowledges AI's potential to "improve lives, advance science and strengthen our economies," but warns that "the rapid development of frontier AI models poses serious safety risks if not properly managed."

The initiative comes just before Thursday's planned meeting in Washington between US President Donald Trump and Chinese President Xi Jinping, the leaders of the two nations at the forefront of AI development worldwide. Notably, neither the United States nor China is among the declaration's signatories.

One of the document's main backers, Finnish President Alexander Stubb, told POLITICO that his aim is to build "momentum" to bring more players on board, including Washington and Beijing. "I think it will be in the interest of both China and the United States to have some kind of guardrails. Our message here is that we need human intervention: humans must be in charge," he said in an interview at Finland's consulate in New York.

The declaration cites as an example of risk the recent cyberattack carried out by "rogue" AI agents against the Hugging Face platform, and from there urges these dangers to be addressed by requiring AI companies to develop "transparent safety protocols," including mandatory pre-deployment testing and independent evaluations conducted by qualified evaluators with sufficient access to assess the risks.

The document stops short, however, of calling for AI development to be slowed or paused: its organizers have preferred to focus on risk management rather than on curbing innovation. Among the measures it does propose are the development and coordination of common standards, greater transparency —including shared reporting of serious safety incidents— and that UN member states "explore the creation of an international institution" to oversee AI, building on existing international mechanisms.

Stubb specified that his personal preference would be to create a new institution, similar to the International Atomic Energy Agency, within the United Nations framework. Even so, he qualified, "we have to leave it open, so that people don't reject it just because it would be within the UN framework," in a clear allusion to Trump's skepticism toward the multilateral body.

Norwegian Prime Minister Jonas Gahr Støre, who promoted the declaration alongside Stubb and also took part in the video interview, said they would welcome both the G7 and the United Kingdom's upcoming G20 presidency taking up the issue: "The UN is big, but the G7 and the G20 are a bit more focused," he said.

The United Kingdom itself, however, has not yet joined the declaration. Among the large European countries, German Chancellor Friedrich Merz backs the initiative, while the signatures of France, Italy and Poland are notably absent.

Stubb stressed that neither he nor Støre are "doomsayers" about artificial intelligence, but he listed three reasons why the lack of effective oversight is worrying: bioweapons created with AI, cyberattacks executed by AI agents, and killer robots that use AI for lethal attacks. At the same time, he insisted that the technology's positive potential can be "much greater" if used properly, citing as examples the eradication of diseases, productivity gains and access to information, among other benefits.