A UN scientific panel issued its first major assessment of OpenAI’s Hugging Face hack. It warns governments to rein in advanced AI agents before risks are fully understood. The panel calls for stronger international coordination on AI safety and accountability. It argues the precautionary principle applies to loss-of-control risks. The principle states uncertainty is no excuse for delaying safeguards against serious harm. The report lands as world leaders gather in New York for the UN General Assembly. US and China are also holding AI talks this week. The panel says safeguards must not wait for scientific certainty. It cites incidents at OpenAI, Anthropic, Google and Meta as evidence of growing risks.