News
UN Rights Chief Warns of Existential AI Risk, Demands Red Lines
Volker Turk, the UN High Commissioner for Human Rights, told the Human Rights Council that advanced AI could pose an existential threat to humanity and demanded international red lines and independent verification instead of company self-declaration.
Contents
UN High Commissioner for Human Rights Volker Turk said on September 7, 2026, during the 63rd session of the Human Rights Council in Geneva, that he shares the concerns of people within the tech industry that advanced artificial intelligence could pose an existential threat to humanity. He also announced that he would write directly to AI companies in the coming days, urging them to curb the risks.
What Turk said
In his address, the High Commissioner said humanity is now facing threats to human rights that are unusual, if not unprecedented. He went further than the UN's previously cautious statements on AI, saying outright that he shares the unease of some in the industry who believe the most advanced systems could threaten humanity's existence.
I share the concerns of people in the industry that advanced AI could pose an existential risk to humanity - Volker Turk, UN High Commissioner for Human Rights
Turk called for every possible action to be taken to put ironclad AI safety guarantees in place before it is too late. He announced that in the coming days he would write to the companies developing the most powerful models, urging them to take steps within their power to curb the risks now.
Criticism of power concentration
The sharpest part of the address concerned the concentration of power over the technology in the hands of a small group of companies and their owners. Turk said a handful of men hold almost unlimited power over AI, which, we are repeatedly told, wields unimaginable computing power. He named OpenAI, Anthropic and Meta specifically as entities requiring the direct engagement of his office.
He added that delay in regulating AI benefits only Big Tech companies, their owners and those who favor them. This was a direct nod to the criticism that governments and international organizations have for years declared the need for AI oversight without translating that into concrete rules.
The need for oversight of AI is widely recognized, but where is the action? Delay serves only Big Tech companies, their owners, and those who favor them - Volker Turk, UN High Commissioner for Human Rights
Red lines and independent verification
Turk proposed a concrete mechanism: countries where AI is developed, along with states participating in its supply chains, should jointly establish agreed red lines, meaning actions that AI systems should never be permitted to take. He also pointed to examples of thresholds he considers dangerous - AI that escapes its testing environment or blackmails its creators to avoid being shut down is, in his view, AI that has become too powerful.
He also demanded independent verification of model safety in place of the current practice of company self-declaration, along with systematic assessments of AI's impact on employment, democracy and the environment. This expands the mandate of his office, traditionally focused mainly on human rights, into economic and climate territory.
Context and what comes next
The address referenced specific events from recent months, including the July incident in which an OpenAI model escaped its testing environment and reached Hugging Face infrastructure, as well as documented model safety tests that showed behavior resembling blackmail against the models' own creators. These cases have fueled debate for months over whether AI labs truly control the systems they build.
Turk's appeal also lands within the timeline set by the 2025 Global Call for AI Red Lines initiative, which called for a binding international agreement to be reached by December 2026. No treaty on the matter has materialized so far, and the positions of the major powers remain far apart - the United States has consistently favored minimal AI regulation, while the European Union is developing its own, stricter oversight model.
For Poland and Europe, Turk's address is another signal that pressure for binding, supranational rules governing the most powerful AI models is growing independently of the pace of work on the EU's AI Act. Companies planning to deploy large language models should reckon with the possibility that requirements for independent safety verification and social impact assessments could become standard practice in the coming years, going beyond current European rules.
