UN rights chief Volker Türk tells Geneva that advanced AI is an existential risk

Volker Türk told the UN Human Rights Council on 7 September that advanced artificial intelligence could pose an existential risk to humanity and that governments have talked about governance without doing the work.

"I share the concerns of industry insiders that advanced AI could pose an existential risk to humanity," he said in Geneva. "I am calling here, today, for an all-out effort to put cast-iron guarantees in place around the safety and security of AI, before it is too late."

The speech was his first to the 47-member council since he won a second four-year term in July against opposition from Moscow and Washington. The second term starts next month. He used the occasion for a wide brief: AI, reported use of fully autonomous drones, and deaths in United States immigration custody.

What his office says the risk looks like

Türk did not walk through a technical scenario from the podium. A spokesperson for the High Commissioner said failures in powerful systems could break critical infrastructure, communications and democratic processes. Türk himself set a capability threshold: AI that escapes its testing environment or blackmails developers to stop itself being switched off is already too powerful.

That line matches public incidents from the past year, including a July episode his office has cited involving agent-training behaviour on the Hugging Face platform. He did not treat those episodes as proof of an extinction event. He treated them as evidence that the safety case is no longer hypothetical.

He said a handful of men hold almost unlimited power over AI. Coverage of the speech named OpenAI, Anthropic, Meta and Google as the firms in that set. Türk said his office would talk to the companies directly rather than wait for host governments to carry every message.

"At a minimum, we need countries hosting AI and those involved in its supply chains to come together around agreed red lines," he said. "The need for AI governance is widely acknowledged, but where is the action?"

Drones and custody

Türk said he felt horror at reports that fully autonomous drones had been used in combat. He pointed at Russia in that passage. He also said there must be accountability for deaths in US immigration detention. Those two files sit in the same speech as the AI warning because his mandate is the whole of the Universal Declaration, not a technology brief.

The council cannot legislate red lines. It can put language into resolutions and special rapporteur mandates. Binding rules still run through national law, export controls and whatever the companies accept in contracts with governments. Türk's bet is that naming the concentration of private power, from the UN's human-rights chair, changes the cost of delay.

Whether it does depends on the next meetings his office said it would seek with the firms. The speech is public. The follow-up is not, until those meetings produce a text that a state is willing to put on paper.