The UN’s secretary general, António Guterres, has warned governments and other key stakeholders that AI development is outpacing regulation and called for global standards on AI safety.

In an address to attendees of the UN’s first Global Dialogue on Artificial Intelligence Governance, Guterres called for better guardrails and a faster public sector response to AI.

“A technology that can reshape economies, transform the world of work, sway elections, and tilt the balance of security is being deployed faster than anyone – including the people building it – can keep up,” he said.

Guterres called on countries to agree on an aligned method to test AI systems for risks, including common baselines for frontier systems, as well international cooperation on responsible AI usage. Without these measures, he added, a “patchwork of incompatible rules raises costs, divides the world – and protects no one”.

The release of powerful AI models such as Anthropic’s Claude Mythos have prompted cross-sector emergency meetings and sparked concerns that AI could soon be used to launch even more sophisticated cyber attacks.

In June, cyber agencies in the Five Eyes intelligence alliance warned that advanced AI attacks may only be months away.

In his address, Guterres said that more must be done to build equitable, publicly-owned AI infrastructure, contrasting the near half a trillion dollars privately invested in AI in 2025 with the “rounding error” of AI for developing countries.

Guterres criticised the extreme energy and water usage associated with AI data centres – calling on major AI companies to sign the UN’s AI Environmental Transparency Initiative and disclose their full carbon, water, and land use – and branded autonomous weapons systems “morally repugnant”.

The UN event is not expected to result in a new treaty but the organisers hope it can steer the decision-making of attending countries, with 40 researchers from around the world set to reveal the findings of a preliminary UN report on the opportunities, risks, and impacts of AI.

---