GENEVA, July 9 - The United Nations agency responsible for digital technologies said on Thursday it will launch a targeted effort to strengthen trust in a new class of autonomous artificial intelligence systems known as AI agents. The International Telecommunication Union (ITU) announced the move during the AI for Good Summit in Geneva as concerns grow over accountability and the limits of human oversight as these systems act more independently.
AI agents are described as a generation of systems designed to operate on users' behalf, performing activities that range from basic scheduling and purchasing to managing more complex business processes. While proponents argue that such systems can boost productivity, the ITU warned they also carry risks, including the potential to impersonate individuals and to take decisions without proper authorization.
To confront those hazards, the ITU said it will establish a Focus Group charged with developing practical frameworks to ensure AI agents remain identifiable, trustworthy and subject to meaningful human control. The work is intended to be especially focused on high-sensitivity areas, with the ITU explicitly citing financial transactions and critical infrastructure as priority domains for safeguards.
"AI agents will soon negotiate, transact and make decisions on our behalf," said Focus Group Co-Chair Debora Comparin. She added that common international foundations were needed to establish who the agents are and how and when they can be trusted.
The planned Focus Group will bring together technical, policy and legal experts. According to the ITU announcement, the group will hold its inaugural meeting in Paris in November and will convene again in Geneva in January.
The ITU framed the initiative as a response to the balance between potential productivity gains and the governance challenges posed by autonomous systems. By assembling multidisciplinary expertise, the Focus Group aims to create shared foundations that define identifiability and human oversight mechanisms for AI agents operating in sensitive and broadly impactful settings.
Dateline: AI for Good Summit, Geneva