Background
<cite index="22-1">On May 8, 2026, China's Cyberspace Administration (CAC), National Development and Reform Commission (NDRC), and Ministry of Industry and Information Technology (MIIT) jointly released the *Implementation Opinions on the Standardized Application and Innovative Development of Intelligent Agents*.</cite> <cite index="22-2">The document marks the first time China has systematically treated agentic artificial intelligence (AI) as a future digital infrastructure and governance object in its own right, rather than simply another application layer built on top of large language models (LLMs).</cite>
A companion instrument arrived weeks earlier. <cite index="10-9,10-10">On April 10, 2026, the CAC and four other departments jointly issued the *Interim Measures for the Administration of AI Anthropomorphic Interactive Services*, which took effect on July 15, 2026.</cite> <cite index="10-11">This is China's first departmental regulation dedicated to anthropomorphic interactive services and forms another key component of the country's AI governance framework, following prior rules on algorithmic recommendation, deep synthesis, and generative AI services.</cite>
Defining the Scope
<cite index="19-4">The *Implementation Opinions* define an AI agent as an "intelligent system capable of autonomous perception, memory, decision-making, interaction, and execution," pulling agents out of the broader "generative AI" bucket regulated by China's 2023 rules and recognising them as systems whose autonomy creates distinct risks and opportunities.</cite>
<cite index="14-6">The anthropomorphic measures, in turn, apply to products and services that use AI technologies to simulate human personality traits, thinking patterns and communication styles, providing emotional interaction with the public in China through text, images, audio, or video.</cite>
Tiered Autonomy and Decision Authorization
<cite index="5-6">Critically, the agent framework defines three categories of agent decision: those reserved exclusively for humans, those permitted with user authorisation, and those agents may take autonomously.</cite> <cite index="3-3">The document states that users "have the right to know and the final decision-making power regarding the autonomous decisions made by the intelligent agent, and that the intelligent agent's actions do not exceed the scope authorized by the user."</cite>
<cite index="5-1,5-2">On safety, the document adopts a tiered governance model, imposing strict filing, testing, and recall requirements on agents in high-risk sectors such as healthcare, finance, transportation, judicial services, and public security.</cite> <cite index="3-7">Lower-risk consumer scenarios are expected to rely more heavily on platform governance, third-party evaluation, and industry self-regulation, supported by a credit evaluation system that can penalise violators.</cite>
<cite index="21-4">The *Implementation Opinions* also lay out 19 specific application scenarios across scientific research, industrial development, consumer spending, public welfare, and governance.</cite>
Anthropomorphic Services: Emotional Boundaries and Minors
<cite index="15-2,15-4,15-5">The *Interim Measures for the Management of Anthropomorphic AI Interactive Services* constitute the only framework globally that explicitly prohibits several activities of anthropomorphic services, and stipulate "emotional boundaries" and the prevention of "overdependence" and the "replacement of social interaction."</cite> <cite index="10-8">The measures introduce new mechanisms for extreme-scenario life intervention, emotional boundary control, and dynamic anti-addiction.</cite>
<cite index="24-4">On July 15, 2026 — the effective date of the *Interim Measures* — ByteDance's Doubao and Alibaba's Qwen pulled their personified companion features, as the measures ban minors from virtual intimate relationships.</cite>
Policy Context and International Comparison
<cite index="5-10">China's 'AI Plus' strategy mandates 70% agent adoption across intelligent terminals and public-sector services by 2027, rising to 90% by a later target date, and Chinese foundational models — principally Qwen and DeepSeek — already rank at the top of international benchmarks for autonomous web traversal, real-world coding, and spreadsheet execution.</cite>
<cite index="24-2">For a US contrast, there is no single federal agency that plays the role China is assigning to the CAC here; oversight of autonomous agents remains fragmented across existing regulators — a gap that is a large part of why the Chinese framework drew international attention.</cite> <cite index="8-3">China had previously made history as the first country to implement binding generative AI regulations with its Interim Measures for Generative AI Services in August 2023.</cite>