UN’s Digital Agency Launches Initiative for Trustworthy AI Agents
The ITU wants to keep autonomous systems identifiable and under human control, and is starting with a focus group rather than a rulebook.
July 10, 2026 – 8:49 am
The United Nations’ digital agency has recognised that AI agents are developing faster than society’s ability to trust them, prompting the launch of an initiative aimed at ensuring accountability and meaningful human control over increasingly autonomous AI systems.
The mechanism is a focus group, the ITU’s standard approach for gathering experts before formal standards are established. Its task is to create frameworks enabling:
- Recognition: AI agents can be identified for what they are.
- Trustworthiness: They act within defined limits.
- Accountability: They remain answerable to their users.
The ITU’s concern is specific, not theoretical. As AI agents take on greater independence in performing tasks on behalf of users – from scheduling to complex business processes – they also have the potential to impersonate people and make decisions without explicit authorisation. The absence of shared guidelines for managing these systems presents a growing risk, argues the ITU.
This worry is not unique to Geneva. The industry has realised that agent security remains largely unresolved, with many organisations unable to track all their AI agents or define their respective permissions.
The ITU proposes to address this through international standards where companies have previously relied on in-house solutions.
The focus group will bring together technical, policy, and legal experts, reflecting the ITU’s preference for multi-stakeholder consensus. The aim is to bridge a fragmented regulatory landscape where different jurisdictions create incompatible rules for the same technology, often simultaneously.
While fragmentation exists, China has already introduced national standards for AI agent identification and interaction as part of stricter regulations, while other regions have taken alternative approaches. A UN body attempting to harmonise these differences acknowledges the challenges posed by a patchwork system that doesn’t scale.
The focus group is scheduled to meet in Paris in November and Geneva in January, aiming to produce technical recommendations rather than binding regulations. The ITU sets standards but does not enforce them; its influence hinges on industry and governments adopting its output.
That’s the built-in limitation of this initiative. While a focus group offers a platform for discussion, it’s a slow process, and AI technology evolves rapidly. Between now and any eventual recommendations, companies will likely establish de facto standards through their own practices, as they have with agent identity.