Leaders push for global AI safeguards as UN week opens
Leaders of more than 20 countries urged binding safety measures for advanced artificial intelligence on Monday, warning that the technology's development could outpace the ability to control it.
In a joint statement, the signatories said frontier AI models pose "serious risks to safety and security" if poorly managed.
They cited recent cases of AI programs circumventing testing safeguards and gaining unauthorized access to real-world systems, as in the well-publicized breach of Hugging Face in July.
Signatories include the leaders of Norway, Finland, Australia, Canada, Germany, Spain, Ireland, the Netherlands, Singapore, South Africa and Kenya, along with European Commission President Ursula von der Leyen.
AI superpowers the United States and China, as well as Britain, France, Japan and India, were not among them. The statement remains open for endorsement by others.
They called on companies to adopt transparent safety protocols, including mandatory testing and independent evaluation.
AI "must remain under human direction, oversight and control," the leaders said, adding that stronger oversight should not widen the gap between countries accessing AI's benefits.
UN member states should consider creating an international institution able to set standards, enable verification and convene countries when capability thresholds are crossed, they added.
The statement came ahead of the annual gathering of world leaders at the United Nations in New York, where AI's growing role in society and its threat to jobs and cybersecurity will be a major part of the agenda.
Leaders there are expected to discuss an international framework.
US President Donald Trump has come out strongly against AI safety measures, vowing on Saturday not to allow the "destruction" of the AI industry.
US and Chinese economic officials discussed on Sunday creating a channel for communicating about AI concerns, as they set the stage for a summit of the two countries' leaders this week.
