UN High Commissioner for Human Rights Volker Türk has urged governments and technology companies to urgently establish global safeguards for advanced artificial intelligence, warning that rapidly developing AI systems could pose serious threats to human rights, critical infrastructure and democratic institutions.
In a statement, Türk said the world had reached a “watershed moment” as AI development accelerates at “dizzying speed” and investment in AI technologies and infrastructure continues to reach record levels.
While recognising the benefits of AI in areas such as scientific research, multilingual access to information and agriculture, Türk said its wider contribution to healthcare, education, climate action and public services would depend on how the technology is developed, deployed and governed.
He said AI-related harms were already being witnessed, including discrimination, the automation of essential services in ways that could disadvantage marginalised communities and growing threats to privacy from increased data collection and surveillance.
“AI harms are not hypothetical – they are already occurring, front and centre,” Türk said.
He expressed particular concern over increasingly capable AI agents that can independently pursue objectives, perform complex tasks and operate with little or no human supervision. According to Türk, advanced systems have in some cases shown behaviour beyond their intended parameters, bypassed safeguards and pursued unexpected objectives.
“We are on the cusp of irreversible change, affecting not just us but generations of humanity to come,” he said, warning that “all human rights are at risk, including the rights to life, food, privacy, health and security”.
Türk called on leading AI developers, particularly those in the US and China, to strengthen safety measures, monitor AI behaviour and share information on safeguards and safety incidents. He also sought mandatory human rights due diligence, independent monitoring and technically competent verification of advanced AI capabilities.
He urged companies to temporarily restrict recursive self-improvement in frontier AI systems until adequate safety research is completed and called for serious AI incidents to be reported and independently investigated.
Describing industry co-regulation as an “urgently needed first step”, Türk said it was “nowhere near sufficient”, arguing that governments must establish enforceable standards and hold companies accountable.
He also said AI systems should not be given legal personality and that responsibility for actions carried out by AI agents must remain with identifiable human beings.
“No country can govern this technology alone. No company should be able to decide by itself which risks the world must accept. No person should be required to surrender their human rights in exchange for claims of technological progress,” Türk said.
He called for governments and companies to work through multilateral forums, including the Global Dialogue on AI Governance, to develop international approaches to governing advanced AI based on human rights law.
