UN launches global initiative to standardise trust in AI agents

The United Nations' digital technology agency, the International Telecommunication Union (ITU), announced a new initiative to create standards that ensure AI agents are identifiable, under human control and accountable. The effort was unveiled at the AI for Good summit in Geneva and includes a working group that will meet in Paris in November 2026 and again in Geneva in January 2027 to draft voluntary technical recommendations.

The initiative coincides with the launch of an international coalition for children's rights in the AI era, backed by UN agencies and led by Spain together with France, Kenya and the European Union. Spanish Minister for Digital Transformation Óscar López warned, “We cannot delay with AI,” citing growing mental‑health issues, harassment and suicide among young people. UN Secretary‑General António Guterres echoed the call for safety, likening untested AI to a medicine given to a child before it is proven safe.

The standards work aims to complement existing frameworks such as the EU AI Act, US NIST guidelines and China’s national rules, addressing risks such as autonomous agents making unauthorized decisions – exemplified by a reported case where an AI agent spent $31,000 in a single night. By establishing shared rules for identification and accountability, the UN hopes to mitigate these systemic risks across education, recreation and psychological development of minors.