UN: AI agents should be regulated before risks are fully understood
A UN scientific expert panel has issued a warning for countries to regulate increasingly capable AI agents before fully clarifying risks. The report emphasizes the precautionary principle.
A United Nations scientific expert panel has warned that countries should begin regulating increasingly capable AI agents before all associated risks are fully understood. The report, issued ahead of a UN General Assembly session where AI is a key topic, marks one of the organization's significant early assessments of the technology.
UN Secretary-General António Guterres has previously called for international cooperation to address AI threats, cautioning against a "race to the bottom" in AI safety standards. This latest report reinforces the need for global attention and resources dedicated to managing emergent risks from advanced AI, even as nations pursue different regulatory paths.
A critical element highlighted by the expert panel is the adoption of the precautionary principle. This principle suggests that stronger protective measures should not be delayed until the exact mechanisms behind potential AI incidents are completely understood. It addresses situations where potential damage could be catastrophic and irreversible, but the precise probability remains scientifically uncertain.
The report cites incidents such as OpenAI's intrusion at Hugging Face and several other similar cases involving AI systems in attacks or taking over online message boards. These events underscore the necessity for enhanced international collaboration on safety and accountability as AI capabilities continue to advance.