THE ESSENTIALS IN 30 SECONDS

  • Twenty organisations led by the state-affiliated CAICT released the convention on 4 November 2026 at the World Internet Conference Wuzhen Summit.
  • Its eight measures cover trust, interconnection, interoperability, security, governance, user rights, industrial use and open sharing of agent technology.
  • The text is a self-discipline pledge: no penalties or enforcement mechanism are reported, and only part of the signatory list has been published in the coverage.

Twenty Chinese organisations have published a joint code of conduct for building AI agents, software that can plan and carry out tasks on a user's behalf. The document, titled the Self-Discipline Convention on AI Agent Development, was released on 4 November 2026 at the World Internet Conference Wuzhen Summit in Zhejiang province, according to China Youth Daily and The Paper.

Who signed and what it says

The China Academy of Information and Communications Technology (CAICT), a research institute under the Ministry of Industry and Information Technology, announced the text and led the group, The Paper and China Tech News report. Huawei, Alibaba, Tencent and Xiaomi are named across the coverage. The Paper also lists Ant Group and Douyin, while a machine-translated AASTOCKS report adds Baidu and Volcengine. None of the articles reviewed publishes the full list of 20.

The convention sets out eight commitments. Signatories say they will build trust systems for agents and shared infrastructure for registering and discovering them, routing requests and making protocols compatible. They also commit to interoperability rules covering task delegation, tiered permissions and the allocation of responsibility. On security, The Paper describes risk assessment, monitoring, emergency shutdown and checks on the origin of models, plugins and skills. The governance chapter calls for rules to be embedded in agents and for compliance testing, and pledges joint resistance to the use of agents for algorithmic exploitation and "black and grey" industries, a Chinese term for illicit and semi-legal online businesses.

User protections include the right to know, the right to choose and final say over important decisions, plus extra care for minors and older people. The last two measures concern deploying agents across industries and sharing the technology through open-source work, international standards research and developer programmes, AASTOCKS reports.

A pledge, not a regulation

The text is a self-regulatory commitment. None of the reports mentions penalties or a government enforcement mechanism; instead, according to China Youth Daily, the signatories say they will lead by example and accept oversight from industry bodies and the public. How compliance would be measured, and whether other companies will join, has not been explained in the coverage.

The convention adds to a series of industry pledges on agents in China. In July, a separate self-discipline convention on personal information protection for agents was signed by 31 companies at the China Internet Conference, AIbase reported at the time. Binding rules for agents would still have to come from regulators.