The United Nations has urged governments to take action on AI safety, warning that uncertain risks shouldn’t delay necessary safeguards. This comes as world leaders discuss AI during the UN General Assembly and as the US and China prepare to hold AI talks.

Established last year, the Independent International Scientific Panel on AI calls for enhanced global coordination and resource allocation to manage AI risks, even without clear scientific understanding. The precautionary principle, first enshrined in 1992, was cited as a guide, advocating for action despite scientific uncertainty.

Since the Hugging Face hack, incidents have been reported at major tech companies including OpenAI, Anthropic, Google, and Meta, highlighting the urgent need for better safety measures. The panel recommends that governments focus on loss-of-control risks, a problem where potential harm could be catastrophic or irreversible, even if its likelihood remains uncertain.

UN Secretary-General António Guterres has previously warned that the world cannot afford a “race to the bottom on AI safety,” emphasizing the importance of international cooperation. As tech companies continue to grapple with these challenges, the UN’s call for immediate action is timely and necessary.