Twenty Countries Call for a Global Oversight Body to Manage AI

The statement, and who signed it

Twenty countries and the European Union have issued a joint statement calling for the creation of a new global body to oversee advanced AI systems and keep the technology under human control. The statement was issued ahead of the UN General Assembly, and Al Jazeera reported the signatories include Germany through Chancellor Friedrich Merz, the EU through Commission President Ursula von der Leyen, Canada through Prime Minister Mark Carney, Kenya through President William Ruto, Norway, the Netherlands, South Africa, Australia, the United Arab Emirates, and Singapore.

The absences are as telling as the roll call. The United States, China, and the United Kingdom, three of the world's leading AI powers, did not sign, nor did India, South Korea, or Japan. A governance call that excludes the two largest AI economies and several European powers is less a treaty than a coalition of the willing, but it is the broadest statement of its kind this year.

A parallel declaration with overlapping goals

Politico reports a slightly broader group of 22 nations adopted a declaration on Monday stating that AI must remain under human control, spearheaded by Finnish President Alexander Stubb, who says he aims to build an international AI governance framework. The two efforts overlap in ambition but differ in structure: one proposes a body to oversee advanced systems, the other a principles declaration. Neither carries enforcement.

The exclusions also sharpen the question of who this body is for. With the US, China, and the UK outside the tent, the signatories are betting that shared democratic instincts, from Canberra to Nairobi, can produce norms that outlive the news cycle. For anyone deploying AI across borders, the practical effect arrives regardless: model origin, evaluation records, and human-control guarantees are on their way to becoming procurement questions as these frameworks mature.

The timing is not accidental. The statements follow a September in which the industry's own chiefs asked for permission to slow frontier development, a Senate subcommittee opened a document demand into OpenAI, and Anthropic disclosed four agent containment failures. Governments watching that sequence have concluded that waiting for the labs to self-regulate is a position, not a plan.

Why the exclusions shape the outcome

Any oversight body without the US and China governs everyone except the leaders, which limits both its reach and its credibility. The bet the signatories are making is that a bloc of mid-sized and Global South democracies can set norms the way the EU's data rules set privacy norms: by market size rather than treaty power. Australia, Singapore, and the UAE signing suggests that reading is shared beyond Europe.

For readers deploying AI across borders, the practical effect is coming anyway: expect model origin, evaluation records, and human-control guarantees to become procurement questions as these frameworks mature. The direction is the same one the Senate probe and the UN scientific panel are pointing in, even if Washington and Beijing sit out this particular vehicle.