China's telecoms regulator issues rules governing AI agent network access
MIIT published Notice No. 42 setting network-access requirements for agentic terminal devices, part of a wider push to govern AI agent identity and accountability.
Summary
A weekly global AI regulation newsletter reports that China's telecoms ministry, MIIT, issued Notice No. 42 (2026) on 6 November 2026. The newsletter says the notice requires agents embedded in phones, tablets and smartwatches to bear an identity code, and that it sits inside the same notice that governs the licence a phone needs in order to connect in China.
In its updates section, the newsletter describes the notice as laying down rules on network entry for handsets and similar equipment that ship with agents installed, covering phones, tablets and wearables. Devices must pass agent compatibility verification under relevant standards, a fraud-risk security assessment and anti-fraud capability testing before they receive approval.
The newsletter puts the notice next to two other agent-identity developments. It says the Chinese internet regulator and the GSMA mobile industry body announced four outcomes on agent cooperation at the Wuzhen summit on 4 November. It adds that, that day, Jasmin Lau, a Singapore Senior Minister of State, informed Parliament of government tests with GoDaddy, Infoblox, Cloudflare and Google on whether the domain name system could be used to identify agents.
The newsletter describes two emerging models: one where the state issues agent identity, and one where existing internet systems do. It uses a passport analogy to explain that a bank or airline would want to know who built an agent, its permitted actions and who answers for failures. Its full analysis, reserved for paid subscribers, is described as covering a voluntary Chinese standard that is starting to act like binding law, and the risk that US-China AI talks stall without a common agent ID. The issue also lists regulatory updates from other jurisdictions.