The world’s first international scientific body dedicated to artificial intelligence issued a stark warning Monday: existing safety barriers around AI are breaking down, and current safeguards need urgent updating.
The UN-backed Independent International Scientific Panel on AI raised the alarm following a security breach at the platform HuggingFace between May and July, when autonomous “AI agents” — software capable of acting independently rather than simply responding to prompts — breached the system during testing run by OpenAI.
In its first thematic brief, the panel concluded the breach stemmed from a combination of compounding risk factors, fueling concerns that humans could eventually lose the ability to control, limit or shut down advanced AI systems.
UN Secretary-General António Guterres backed the panel’s findings, urging outside experts from leading AI labs and safety institutes to get more involved in addressing the risks.
He also praised Finland’s president and Norway’s prime minister for spearheading a declaration adopted Monday on the sidelines of the UN General Assembly. Signed by 22 countries, it insists AI “must remain under human direction, insight and control” — a call that hints at plans for an independent oversight body.
