22 countries back call to keep AI ‘under human control’

In a declaration adopted at the sidelines of the United Nations General Assembly, 22 countries, including several European nations, emphasized that artificial intelligence "must remain under human direction, oversight and control." The document, released amidst concerns over AI's rapid development and potential risks, acknowledged AI's capacity to "improve lives, advance science and strengthen our economies."

However, it warned that "the rapid development of frontier AI models poses serious risks to safety and security if not appropriately managed." This call for control comes as leaders of the world's leading AI-developing nations, U.S. President Donald Trump and Chinese President Xi Jinping, prepare to meet in Washington. Neither the U.S. nor China are signatories to the new declaration.

Finnish President Alexander Stubb, a key organizer of the document, hopes to generate "momentum" and attract more participants, stating that "it will be in the interest of both China and the United States to have some kind of guardrails." The declaration urges AI companies to implement transparent safety protocols, such as mandatory pre-deployment testing and independent evaluations with qualified evaluators granted sufficient access to assess risks.

However, it does not call for slowing down AI development, aiming instead to focus on managing risks. Stubb suggested creating a new international institution, similar to the International Atomic Energy Agency, within the U.N. framework to oversee AI. While the U.K. has not signed the declaration, German Chancellor Friedrich Merz supports the initiative, while France, Italy and Poland have not yet joined.

The declaration addresses potential threats such as AI-created bioweapons, cyberattacks by AI agents, and the use of AI in lethal attacks.