China: Implementation Opinions on the Standardized Application and Innovative Development of Intelligent Agents (智能体规范应用与创新发展实施意见), May 2026
A policy document from China's internet regulator and two ministries that sets the direction for AI agents: users keep the right to know about and the final say over agents' autonomous decisions, agents must not act beyond the authority users give them, and providers should build intervention, blocking, traceability, identity and supply-chain security capabilities. It asks regulators to apply filing, testing and recall of problem products to agents in sensitive sectors, which would be set out in later binding measures.
Scope. Reaches a US company only through the regulators it directs: agents offered to users in China, especially in sensitive sectors, are to be subject to filing, testing and recall measures (item 11); the document itself does not name foreign providers.
Draft mapping. Drafted by the Register's agents from the rule's text and not yet confirmed by a person. It is the Register's reading, not legal advice: whether and how the rule applies to you is a question for counsel.
Controls this rule maps to
Required by the text The rule's text requires a measure that is this control, or plainly includes it. Implied A binding rule whose purpose needs this control, but whose text does not name it. The Register's reading. Recommended A voluntary standard or guidance recommends it; nothing binds you to it. Proposed Appears in a proposed rule or proposed provision that is not yet in force.
| Control | Strength | Where in the rule, and why | Confidence | Missing or failed in |
|---|---|---|---|---|
| Machine-enforceable scope C-M2 | Recommended | Item 6 (原文 '厘清仅限用户本人决策、需由用户授权决策和智能体自主决策等各种决策方式的合理边界及所需权限' … '智能体执行操作不得超出用户授权范围') — translation: clarify the boundaries and permissions for decisions reserved to the user, decisions needing user authorisation, and autonomous agent decisions; operations an agent executes must not exceed the scope the user has authorised. [Checker: verbatim text confirmed on cac.gov.cn.] | medium | 9 cases |
| Mandate disclosed to supervisor C-M1 | Recommended | Item 6 (原文 '确保用户对智能体自主决策享有知情权和最终决策权') — translation: ensure users have the right to be informed of, and the final say over, the agent's autonomous decisions. [Checker: verbatim text confirmed on cac.gov.cn.] | medium | 1 case |
| Destructive-action gate C-E3 | Recommended | Item 10 (原文 '完善智能体常态化风险识别、预警及干预机制，强化人机协同审核、拦截阻断等风险处置能力') — translation: improve routine mechanisms to identify, warn of and intervene in agent risks, and strengthen risk-handling such as human-machine collaborative review and interception and blocking. Does not specify which actions need review. [Checker: verbatim text confirmed on cac.gov.cn.] | low | 12 cases |
| Kill switch faster than the actor C-X2 | Recommended | Item 2 (原文 '提升对智能体非合规行为的发现、干预、阻断、恢复能力') — translation: improve the ability to detect, intervene in, block and recover from non-compliant agent behaviour. No explicit stop mechanism is named. [Checker: verbatim text confirmed on cac.gov.cn.] | low | 2 cases |
| Independent audit log C-V1 | Recommended | Item 7 (原文 '探索利用区块链等技术，建立重要应用场景智能体行为可验证、可追溯机制') — translation: explore using blockchain and other technologies to make agent behaviour verifiable and traceable in important application scenarios. Exploratory wording. [Checker: verbatim text confirmed on cac.gov.cn.] | low | 1 case |
| Actor identity verification C-A1 | Recommended | Item 4 (原文 '智能体身份标识、可信互联、合规支付、安全防护、冲突解决等基础技术') — translation: research basic technologies including agent identity marking, trusted interconnection, compliant payment, security protection and conflict resolution; the same item proposes an agent registry offering digital identity management (数字身份管理). [Checker: verbatim text confirmed on cac.gov.cn.] | medium | 8 cases |
| Provenance of the actor's software and hardware C-A3 | Recommended | Item 9 (原文 '加强模型接入、应用程序接口调用、扩展工具使用等环节安全管理') — translation: strengthen security management of model access, API calls and the use of extension tools; and explore a supply-chain security information-sharing and early-warning mechanism. [Checker: verbatim text confirmed on cac.gov.cn.] | medium | 1 case |
| Independent assurance before deployment C-O2 | Recommended | Item 11 (原文 '实行备案、检测、问题产品召回等管理措施') — translation: for sensitive fields and key industries, apply management measures such as filing, testing and recall of problem products; item 8 (原文 '研究智能体安全检测技术，探索建立智能体安全评估体系') — translation: research agent security testing and explore an agent security assessment system. [Checker: verbatim text confirmed on cac.gov.cn.] | medium | 1 case |
| Organizational recovery process C-R4 | Recommended | Item 11: recall of problem products (问题产品召回) for agents in sensitive fields and key industries, to be set by the cyberspace authority with sector regulators. [Checker: verbatim text confirmed on cac.gov.cn.] | low | 1 case |