UN scientific panel urges precautionary safeguards for advanced AI agents

A United Nations scientific panel has urged governments to adopt safeguards for increasingly capable AI agents without waiting for researchers to resolve every uncertainty about their risks. Its first thematic brief calls for more resources, international coordination and stronger safety and accountability measures. The panel applies the precautionary principle to potential loss-of-control incidents, arguing that uncertain likelihood should not delay action where harm could be catastrophic or irreversible. The recommendations arrive as the United States and China prepare for AI discussions, but they do not create binding rules.