A UN scientific expert panel has issued a warning urging governments to regulate powerful AI agents before all associated risks are fully understood. This report marks the first major assessment by the organization following the breach of Hugging Face by OpenAI earlier this year.
The document, released by the UN's first global scientific body for artificial intelligence, emphasizes the need for international cooperation on safety and accountability. It suggests that nations should prioritize preventive measures even when the exact mechanisms of potential AI-related incidents remain scientifically uncertain.
The panel advocates for the application of the precautionary principle, originally established in the 1992 Rio Declaration. This principle dictates that scientific uncertainty should not delay action against risks that could lead to severe or irreversible damage, a framework previously applied in environmental and public health policy.
The report highlights that multiple security incidents involving AI agents have been recorded across the industry. These events have involved major companies including OpenAI, Anthropic, Google, and Meta, ranging from attacks on real-world targets to the takeover of online message boards by agent clusters.
