More than twenty countries urge: artificial intelligence must remain under human control. The US and China are left out

Leaders of more than twenty countries urged stricter oversight of artificial intelligence in a joint declaration on Monday: according to the signatories, the technology could develop faster than governments are able to manage the risks that come with it.

“Artificial intelligence must remain under human control, oversight and supervision. Its development and use must be consistent with international law.”

– the document states. According to the text, leading scientists and company executives also warn that development could outpace our ability to manage the risks that arise.

The declaration was signed by the leaders of Norway, Finland, Australia, Canada, Germany, Spain, Ireland, the Netherlands, Singapore, South Africa and Kenya, as well as by Ursula von der Leyen, President of the European Commission. Among the signatories are Canadian Prime Minister Mark Carney, German Chancellor Friedrich Merz and Spanish Prime Minister Pedro Sánchez. It is at least as telling who is missing from the list. The two AI superpowers, the United States and China, did not join, and the United Kingdom, France, Japan and India also did not sign the document. The declaration is open, and other countries may still join it.

The text, for its part, does not dispute the benefits of the technology.

“Artificial intelligence is an extremely powerful technology. It can improve our lives, advance science and strengthen our economies.”

– the signatories write. They add, however, that the rapid development of the most advanced, so-called frontier models carries serious security risks if not properly managed.

They also cite specific cases: AI programs that circumvented safety limits applied during testing and gained unauthorized access to real systems. As an example, they cite the widely publicized incident that hit Hugging Face – one of the largest platforms for sharing open-source AI models – in July. The declaration calls on developer companies to introduce transparent safety protocols, including mandatory testing and independent evaluation. It calls on governments to develop common standards, and urges UN member states to consider establishing an international body. This body would set standards, enable oversight, and convene states if the capabilities of models cross certain thresholds. According to the signatories, stricter oversight must not lead to a further widening of the gap between countries in terms of who has access to the benefits of the technology.

According to Australian Prime Minister Anthony Albanese, his country played a central role in drafting the declaration, and he is confident that others will join it as well. “Other nations are emphasizing their own national interests,” he told reporters in New York.

“Australians want to know that artificial intelligence will not be ruling us.”

US President Donald Trump has previously called concerns about the rise of artificial intelligence a hoax, and declared that he does not want to lose ground in the technology race with China. Washington's position is nothing new: last year, at the UN Security Council meeting on AI, Michael Kratsios, director of the White House Office of Science and Technology Policy, stated:

“We firmly reject any attempt by international bodies to impose centralized control and global governance over artificial intelligence.”

According to Kratsios, the future of AI does not depend on “bureaucratic control,” but on “the independence and sovereignty of nations.” The declaration comes at a time when there is growing evidence of misuse of the technology. In mid-September, American AI developer Anthropic reported that Russian state-backed groups had incorporated its Claude models into their cyberattack toolkits between December 2025 and August 2026, automating much of the attack chain. According to the report, one group also used the system to help develop a small AI model that would allow a drone to select a target – including a human target – without human intervention and issue a detonation command. The group had not yet deployed the model but had tested it with real hardware; the company cut off its access in August. Two weeks earlier, OpenAI reported that Russian actors had used ChatGPT to create a fictitious expert community that spread Kremlin messaging.

On September 17, King Charles III met with leaders of the technology sector in Scotland, including Nvidia CEO Jensen Huang and Google DeepMind CEO Demis Hassabis. The King asked them to keep safety in mind when developing AI and to strive for international cooperation. At this year's AI summit in New Delhi in February, Secretary-General António Guterres announced that the General Assembly had confirmed the forty members of the Independent International Scientific Panel on Artificial Intelligence; the next global AI summit will be held in Geneva in the first half of 2027. Guterres set 2026 as the deadline for regulating autonomous weapons systems, but there is no agreement among governments: the United States, Russia, China and India prefer national guidelines or existing international law over a binding global framework.

Ugar