By Hamza Faisal
China has moved to give the AI agents living inside smartphones an identity of their own. The country's telecoms regulator has issued new rules requiring devices that ship with a pre-installed AI agent to carry an identity code for that agent, and to pass compatibility and anti-fraud checks, before they are allowed onto a mobile network.
The Ministry of Industry and Information Technology (MIIT) published the rules as Notice No. 42 of 2026 on November 6. The notice folds so-called "agent terminals" — phones, tablets and wearable devices with an AI agent built in — into China's existing network-access management system, the licensing system every phone must clear before connecting in China.
Under the notice, a device cannot win network-access approval until its built-in agent has undergone compatibility verification, a fraud-risk security assessment and advance anti-fraud capability testing. Developers must confirm their agents work with the specific terminal models and operating systems they ship on, and product settings should disclose which devices have completed the adaptation checks.
An identity code for every agent
The most striking element is the identity requirement. Each agent built into a phone, tablet or smartwatch must carry an identity code, a move analysts have described as a "passport for bots" — a way for networks and regulators to know exactly which agent is acting on a device, and on whose behalf.
The rules also set expectations on how agents use permissions, handle communications functions, protect network and data security, and respect user rights. The aim is to make sure agents operate stably and users' rights are protected as agentic features spread into everyday devices.
The notice, signed on October 24, is a departmental rule inside telecoms network-access management rather than a sweeping new AI law. Even so, it is one of the first concrete rules anywhere to treat the AI agent itself — not just the hardware it runs on — as something that must be identified and tested before it reaches consumers.
Part of a wider clampdown on agents
The move comes as Beijing tightens oversight of autonomous AI software more broadly. China's 15th Five-Year Plan for informatisation, published in late October, calls for better filing and safety-risk assessment of agents, on-device large models and services that mimic human interaction, while a mandatory agent safety standard remains in drafting.
That scrutiny has already produced separate warnings. Bloomberg reported on November 6 that Chinese authorities moved to restrict state-run enterprises and government agencies from running OpenClaw, a popular agentic AI tool, on office computers, citing security risks from software that holds broad access to private data and can communicate externally.
For phone makers, it is a new gate in the launch process: an AI feature that once shipped as a software update may now need its agent verified, tested for fraud risk and registered before the handset can legally connect.
What it means for the rest of the world
Other regulators are watching similar questions, but none has yet tied agent identity to network access in this way. The European Union's AI rules and the United States' patchwork of state laws focus mainly on what models can do and how their risks are disclosed, not on giving each agent on a device a registered identity before it can go online.
China's approach will be closely watched by handset makers and agent developers selling into the world's largest smartphone market. If the model works, other countries may borrow the idea. For now, the message from Beijing is clear: the age of the anonymous AI agent is ending, at least on Chinese phones.



