Key facts
- 22 countries adopted a declaration calling for AI to remain under human control.
- The declaration was adopted during the UN General Assembly.
- It warns of serious risks to safety and security from frontier AI models if not managed.
- The document urges AI companies to implement transparent safety protocols and independent evaluations.
- It suggests exploring the creation of an international institution to oversee AI.
- Finnish President Alexander Stubb is a key proponent of the initiative.
Twenty-two nations, predominantly European, have adopted a declaration asserting that artificial intelligence must be kept under human direction, oversight, and control. The declaration, made during the United Nations General Assembly, acknowledges AI's potential benefits for improving lives, advancing science, and strengthening economies, but also highlights the significant safety and security risks posed by rapidly developing frontier AI models if not properly managed.
Key proponents of the initiative, Finnish President Alexander Stubb and Norwegian Prime Minister Jonas Gahr Støre, aim to build momentum and encourage broader international participation, including from the U.S. and China, who are not signatories to the current declaration. Stubb expressed a personal preference for creating a new UN-framework institution, similar to the International Atomic Energy Agency, to oversee AI, while acknowledging the need for flexibility to gain wider acceptance.
The declaration calls for AI companies to develop transparent safety protocols, including mandatory pre-deployment testing and independent evaluations by qualified experts. It also advocates for the development and coordination of common standards and enhanced transparency, such as shared reporting of serious safety incidents. The initiative suggests leveraging existing international mechanisms and exploring the establishment of a new global body for AI oversight.
While the document does not call for a slowdown or pause in AI development, it emphasizes managing risks. Prime Minister Støre indicated that the G7 and the upcoming UK presidency of the G20 could play a role in advancing these goals. German Chancellor Friedrich Merz supports the initiative, though signatures from France, Italy, and Poland are notably absent. Presidents Stubb and Støre emphasized they are not "doomsdayers" but cited AI-created bioweapons, cyberattacks by AI agents, and autonomous lethal weapons as significant threats requiring effective oversight.