UN agency launches group to make AI agents trustworthy

The United Nations' agency for digital technologies has launched a new initiative aimed at making autonomous AI systems trustworthy and accountable, as AI agents increasingly take on tasks once reserved for humans.

The International Telecommunication Union announced the ITU Focus Group on Trust and Identity for Humans and Agentic AI on Thursday at the AI for Good Global Summit in Geneva.

AI agents differ from earlier generative AI tools in that they act independently on a user's behalf, handling everything from scheduling and purchasing to complex business workflows.

That independence is exactly what concerns the ITU: agents capable of acting without direct human input also carry the risk of impersonating people or organisations and making unauthorised decisions, particularly in sensitive areas like financial transactions or critical infrastructure.

The focus group will convene a variety of experts with technical, policy, and legal expertise in order to develop mechanisms for maintaining the identification, trustworthiness, and human-in-the-loop control of AI agents.

"As AI agents will very soon negotiate, transact and make decisions on behalf of people," explained Co-Chair Debora Comparin, "a common basis of international understanding is necessary to know who these agents are and whether we can trust them."

As Amir Banifatemi, the other co-chair, put it: "It is about basic accountability – identity means 'who' and trustworthiness means 'how' the actor is behaving."

The work of the focus group is a part of a bigger theme of this year's ITU summit as stated by ITU Secretary General Doreen Bogdan-Martin.