Zetrix AI and CAICT launch service to identify and govern AI agents

A digital identity framework for AI agents

The service was announced at the World Internet Conference in Wuzhen, China, and the ICT Development Forum 2026 in Dubai, United Arab Emirates. It is intended for businesses, governments and digital service providers that need to determine whether an AI agent is authorised to act, who it represents and whether its activity remains within approved limits.

Zetrix AI and CAICT said the service combines verifiable credentials with delegated permissions and checks before protected actions proceed. Unlike conventional digital identity systems that primarily verify human users, the Know Your Agent, or KYA, framework is designed to link an AI agent to an authenticated individual or organisation.

The service builds on W3C Verifiable Credentials and uses AvatarLabs' Agentic Wallet technology to manage credentials, verify delegated permissions and maintain evidence of agent activity.

Permissions can define what an agent is allowed to do

Under the proposed framework, an owner can specify which services an agent may access, what information it can use, who it can transact with and how much it can spend. A business could, for example, allow an agent to make payments to approved suppliers within a defined limit while requiring human approval for larger transactions.

The system is also designed to require an agent to demonstrate its identity and authorisation before accessing protected government or enterprise information. If one AI agent delegates a task to another, the delegated authority cannot exceed the permissions originally granted. Permissions can be revoked, and signed activity records are intended to provide an auditable trail.

TS Wong, Group Managing Director of Zetrix AI, said: “The question is no longer just what AI can do, but who is responsible for what it does.” He added that the company is working toward a system in which “every AI agent can be identified, every permission can be verified and every action can be traced to an accountable owner.”

Pilot will test identity across jurisdictions

The planned cross-border pilot will involve participating organisations from the Middle East, Africa, China and ASEAN, subject to confirmation. It will examine potential uses in cross-border trade, digital payments, government services and interactions between autonomous AI agents.

The pilot is expected to assess identity verification, permission enforcement, revocation mechanisms, privacy safeguards and accountability records. It will also explore how AI agents can operate across different national identity ecosystems while respecting local regulatory requirements, data protection considerations and jurisdictional control.

According to CAICT, the joint service is intended to help create “a cross-border trusted network enabling interoperable Agentic AI around the world.”

Findings may feed into international standards work

Zetrix AI and CAICT plan to submit the pilot's findings to the International Telecommunication Union, IEEE and other Standard Development Organizations for consideration in ongoing work on global AI identity and accountability standards.

More specifically, the organisations plan to jointly submit findings to the ITU's Telecommunication Standardization Sector Study Group 17, through its Focus Group on Trust and Identity for Humans and Agentic AI. The submissions are expected to include implementation findings, cross-border use cases and recommendations covering agent identity verification, delegated authority and accountability, subject to applicable ITU processes.

For users and organisations, the initiative addresses a central issue in autonomous AI: how to allow agents to perform tasks independently while keeping their authority limited, traceable and linked to a responsible owner. The pilot's results could help inform technical discussions about how such controls work across borders, although participation and the proposed applications remain subject to confirmation.