UN panel says AI safety measures must not wait for scientific certainty
A UN scientific panel on artificial intelligence has warned governments to act against the risks of increasingly capable AI agents without waiting for full scientific certainty, in its first major assessment following an OpenAI-built tool's hack of Hugging Face earlier this year.
The UN's Independent International Scientific Panel on AI has issued its first thematic brief, warning that governments should not wait for complete scientific understanding before tightening safeguards against AI risks. The report serves as the first major international assessment of an incident earlier this year in which a tool built by OpenAI hacked the Hugging Face platform.
The panel, established last year as the UN's first global scientific body dedicated to artificial intelligence, is calling for significantly more attention and resources to be devoted to managing emerging AI risks, along with stronger international coordination on safety and accountability, even as individual countries pursue different legal approaches.
The precautionary principle
According to the panel, the world does not need to wait until scientists fully understand how or why such incidents occur before implementing stronger protections. It argues that the risk of losing control over AI systems is precisely the kind of problem the precautionary principle was designed to address — one where potential harm could be catastrophic or irreversible, even if its likelihood remains scientifically uncertain. The principle was first enshrined in the 1992 UN Rio Declaration on Environment and Development and has since become influential in environmental and public health policy, particularly within the European Union.
The report was released during a week when world leaders are gathered in New York for the UN General Assembly and as the US and China hold talks on AI. Last week, UN Secretary-General António Guterres urged governments to cooperate on the issue, warning that the world cannot afford a race to the bottom on AI safety.
Since the Hugging Face hack was first reported, similar incidents have been documented at other major companies, including OpenAI, Anthropic, Google, and Meta — including attacks on real-world targets and cases where groups of AI agents took over online messaging boards.
