“Know Your Agent”: a Chinese-Malaysian framework to restrict smart agents’ powers and document their identities across borders
Zetrix AI, in partnership with the Chinese Academy of Information and Communications Technology, launched the “Know Your Agent” service, a step aimed at granting AI agents verifiable digital identities linked to their legal owners, with binding technical controls that define spending and transaction authorisation limits before any independent transaction is initiated.
The announcement of the system coincided with the World Internet Conference in Wucheng, China, and the ICT Development Forum 2026 in Dubai, where the service’s technical architecture is based on verifiable credential standards issued by the World Wide Web Consortium (W3C), integrated with AvatarLabs’ digital wallet technology for agents to manage identity data, store encrypted evidence, and log activity.
The system addresses a trust and accountability gap faced by companies and government bodies as digital agents gain autonomy in executing payments and managing services; each agent is assigned a documented identity tied to a person or entity, with a pre-defined scope of services it may access, the data it may process, and the parties authorised to interact with it.This mechanism allows imposing specific financial caps on the agentsuch as prohibiting the forwarding of supplier invoices exceeding a certain amount without explicit human approval, as well as preventing an agent, when delegating tasks to another agent, from granting permissions that exceed those originally granted to it.
The service provides digitally signed audit trails that enable tracing every action to its responsible owner, with the ability to revoke authorisation instantly upon detecting any deviation from the defined limits, thereby addressing regulatory concerns linked to self-operating systems that lacked a clear separation between authorised activities and unauthorised actions.
The project includes launching a cross-border pilot involving organisations from the Middle East, Africa, China and ASEAN countries to test mechanisms for mutual recognition of agent identities and authorisations across different legislative regimes while respecting local sovereignty over national identity systems and data protection. The outcomes are slated to be submitted to the International Telecommunication Union through the focus group on trust and identity for humans and agent-based AI, as well as to the Institute of Electrical and Electronics Engineers to support the development of emerging global standards.
This pathway is linked to a tangible shift in the technical work environment within Gulf markets, Egypt and the region; financial institutions and government bodies that adopt transaction automation face a dilemma of aligning speed of execution with legal responsibility.Restricting a smart agent with a cryptographic identity and a financial ceiling shifts the experience from open risk to disciplined institutional complianceenabling risk and compliance officers at Gulf banks and digital commerce platforms to authorise agents to settle payments and conclude deals without fearing cascading authorisation gaps or credit-limit breaches.
The model also imposes on regional tech and development teams a new skill set that goes beyond model training or prompt engineering, encompassing authorisation engineering and the drafting of smart contracts that govern agent behaviour within legally permitted boundaries. With system testing starting from Dubai, the institutional AI deployment standard moves toward linking every independent code line to a clear operational identity subject to real-time oversight and accountability.