On July 15, 2026, China activated the world’s first national governance framework for agentic AI, establishing two binding regulations: the Implementation Opinions on the Standardized Application and Innovative Development of Intelligent Agents and the Interim Measures for the Administration of Anthropomorphic AI Interactive Services. This is not a bolt-on to existing generative AI rules. It is a fundamental shift in how digital infrastructure is governed — one that treats autonomous agents as a distinct category with their own regulatory architecture, guided by a philosophy of deploy first, govern along the way.
The Implementation Opinions, issued jointly by the Cyberspace Administration of China (CAC), the National Development and Reform Commission (NDRC), and the Ministry of Industry and Information Technology (MIIT), define AI agents as “intelligent systems capable of autonomous perception, memory, decision-making, interaction, and execution.” This definition is the foundation for a three-tier decision authorization framework. Developers must categorize agent actions into three distinct buckets: decisions reserved exclusively for users, decisions requiring user authorization, and autonomous agent decisions. The regulation mandates that users must retain “the right to know and the ultimate decision-making authority over autonomous agent actions.” The CAC’s official Q&A frames the risk plainly: “The high autonomy and high permission levels of agents have also created security risks such as privacy leakage, unauthorized operations, and loss of behavioral control.”
This governance structure is tied to China’s push for “indigenous controllability” — a principle embedded throughout the framework that promotes compatibility between agents and open-source chips, domestic operating systems, and foundation models. This is a strategic signal: foreign hardware may be permitted, but the software architecture governing high-stakes decision-making is steered toward Chinese-built stacks. The timing sharpens the contrast. Nvidia’s China market share collapsed to zero by May 2026 under US export controls, while Huawei HiSilicon’s AI chip revenue is projected to reach $12 billion this year — a 60% year-over-year increase.
Beyond individual product compliance, China is building toward an “Intelligent Internet” architecture: agent registration platforms, digital identities for agents, capability declarations, and an Agent Internet Protocol (AIP) to facilitate multi-agent interoperability. The ambition is concrete: 19 identified application scenarios spanning scientific research, industrial development, consumption, public well-being, and social governance, backed by a State Council target of 70% adoption of intelligent agents in smart terminals by 2027. For sensitive sectors — healthcare, finance, transportation, judicial services, public security — the framework requires mandatory filing with regulators, compliance testing, and product recall provisions. Lower-risk sectors are managed through self-assessment and industry self-regulation.
The international comparison reveals the structural gap. The European Union’s AI Act, with its transparency measures taking effect August 2, governs agentic AI only indirectly through high-risk pathways, lacking formal agent-specific provisions. The United States has no federal equivalent — a Congressional Research Service report updated July 6, 2026, confirmed there is no known US government guidance specifically on agentic AI. China, by contrast, has implemented a comprehensive three-part package: the voluntary Ethics-Safety Guidelines for AI Applications 1.0, followed by two binding instruments. The Anthropomorphic Measures target AI services that simulate human personality, thinking patterns, and communication styles, requiring mandatory algorithm filing with the CAC and protections for minors.
This framework reflects Beijing’s focus on maintaining oversight of decision-making logic within its domestic digital ecosystem — not by slowing deployment, but by holding governance strings for the highest-risk tiers. By providing a structured environment with clear adoption targets and 19 application scenarios, China is positioning itself to scale agentic AI at a pace Western regulators have not yet matched. Whether this model can manage the complexities of autonomous decision-making without stifling the innovation it seeks to harness remains an open question — but the structural gap is real, and it is growing.