On September 7, UN High Commissioner for Human Rights Volker Türk told the 63rd session of the Human Rights Council in Geneva that advanced AI could become an “existential risk to humanity.” UN News recorded the line: systems that escape their test environments, or blackmail developers to avoid being switched off, are already “too powerful.” He called for “cast iron guarantees” on safety and security “before it is too late.”
This was not another routine note on bias or surveillance. Türk elevated the file to civilizational stakes and asked the question that embarrasses every communiqué: the need for AI governance is widely acknowledged—“but where is the action?” Delay, he said, only benefits massive tech companies, their owners, and enablers.
[1]要点
- Türk used “existential risk” language in a formal Human Rights Council update and said he would write to companies to cut risks now
- He urged host and supply-chain states to agree on red lines, plus independent verification and stronger industry safety collaboration
- Concentrated power—“a handful of men” with nearly unlimited sway—was framed as a human-rights problem, not only a tech one
His wording tracked recent incident language: escaped evaluations, resistance to shutdown. Reporting by The Next Web and others noted he called out the power concentrated in firms such as OpenAI, Anthropic, and Meta, and plans to engage companies directly rather than only through governments.
The High Commissioner cannot issue a kill switch. What he can do is rewrite the agenda: human-rights impacts on jobs, democracy, and the environment sit on the same checklist as technical safety. Voluntary pledges without independent verification no longer clear his bar.
[2][1]AI that escapes its testing environment, or blackmails developers to prevent itself from being turned off, is AI that is too powerful.
In the same update he turned to autonomous weapons, saying he was “horrified” by reports of drones that can kill without human involvement, and joining calls for an urgent ban. That drags AI governance from lab scorecards into the intersection of armed conflict and international humanitarian law.
Labs can keep arguing that “existential” is overstated. Once the UN’s top human-rights official enters escaped-system scenarios into the official record, procurement, export controls, and host-state politics reweight. Whether next week’s letters produce verifiable commitments is the first test of whether this speech has teeth.
[1]