Finland and Norway lead a call for control of frontier AI

Finnish president Alexander Stubb and Norwegian prime minister Jonas Gahr Støre launched a joint statement at UN headquarters in New York on the eve of the General Assembly’s high-level week, backed by the leaders of about 20 countries and the European Commission — among them Germany, Spain, the Netherlands, Canada, Australia, Singapore, the United Arab Emirates, Kenya and South Africa. It was framed around recent evidence that the most capable systems were slipping their tests:

Capable AI systems circumventing testing safeguards, exploiting vulnerabilities and gaining unauthorized access to real-world systems

— A Call for Control of Frontier AI Models

The statement insisted that “AI must remain under human direction, oversight and control”, and set out asks at three levels. Companies should adopt transparent safety protocols with “mandatory pre-deployment testing and independent evaluation”. Governments should coordinate standards and set up shared incident reporting. And UN member states should create an international institution to “set standards, enable verification, and convene states when capability thresholds are crossed” — while addressing the risks without widening the gap between countries in access to AI’s benefits.

More leaders endorsed it over the following days, among them those of Austria, Romania, Liechtenstein, Luxembourg, France and Sierra Leone, and the statement was left open for others. The two countries that host nearly all frontier development, the United States and China, were absent. The statement arrived in a week crowded with calls for the same thing: the frontier labs’ own pleas to pace development, the European Commission president’s State of the Union remarks, and, two days later, the UN Security Council’s first briefing on AI safety.