July 2026 is shaping up to be a defining month for artificial intelligence governance in China. In a coordinated push to establish clear guardrails for the next generation of AI technologies, Beijing has enacted three major regulatory frameworks that will fundamentally alter how AI is developed, deployed, and interacted with across the country.
These new rules move beyond the broad principles of earlier regulations, targeting specific, highly advanced applications: autonomous AI agents, anthropomorphic chatbots, and the foundational ethics of AI safety. Together, they represent the most comprehensive and granular AI regulatory regime implemented by any major economy to date.
The AI Agent Framework: Moving Beyond Chatbots
The most significant of the three new regulations is the Implementation Opinions on Intelligent Agents, which officially takes effect on July 15, 2026. According to Rimon Law’s China AI Law Brief, which provides a comprehensive analysis of the new frameworks, this is the world’s first dedicated policy specifically targeting autonomous AI agents. 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), the framework covers the full lifecycle of agentic AI deployment.
The regulation defines AI agents as intelligent systems capable of autonomous perception, memory, decision-making, interaction, and execution. This marks a crucial regulatory shift from passive, prompt-response generative AI (like early LLMs) to active, goal-oriented systems that can take actions in the real world or across digital environments.
The framework establishes a differentiated governance model based on risk. AI agents deployed in sensitive sectors, specifically healthcare, transportation, media, and public safety, will face mandatory filing requirements, rigorous compliance testing, and strict product recall provisions if they exhibit harmful autonomous behavior.
Despite strict oversight in high-risk areas, the policy’s overarching goal is to accelerate adoption. The government has set an aggressive target: achieving a 70 percent adoption rate of intelligent agents in smart terminals by 2027. To drive this, the framework identifies 19 specific application scenarios spanning scientific research, advanced manufacturing, and consumer services where the deployment of AI agents will be actively encouraged and supported.
Regulating Virtual Companions and Emotional AI
Also taking effect on July 15 are the Interim Measures for Anthropomorphic AI Interaction Services. This regulation directly addresses the booming market for AI-driven emotional interaction, virtual companions, and chatbots designed to simulate human personalities.
As AI companions become increasingly sophisticated, Beijing is moving to mitigate the psychological and social risks associated with deep human-AI engagement. The new rules impose strict obligations on developers and platforms. All anthropomorphic AI services must undergo security assessments and algorithm filing with the CAC. Crucially, they must clearly and continuously disclose to users that they are interacting with an AI, not a human.
The regulation places a heavy emphasis on user well-being and addiction prevention. Platforms must implement mandatory reminders after 2 hours of continuous use and provide an “unobstructed exit mechanism” to help users disengage. Furthermore, the rules mandate crisis intervention protocols; if an AI detects that a user is showing signs of self-harm or extreme psychological distress, the system must trigger appropriate safety measures.
The protection of minors is a central pillar of the framework. The regulation strictly prohibits providing minors with “virtual intimate relationship services,” including virtual relatives or virtual life partners. Any anthropomorphic service targeting users under 14 requires explicit parental or guardian consent.
Data privacy is also heavily regulated. User interaction data—the deeply personal conversations users have with their virtual companions—cannot be used for further model training without explicit, opt-in consent, and users must be provided with the ability to permanently delete their chat histories.
Establishing the Ethical Baseline: TC260 Guidelines
Preceding the two binding regulations was the release of the TC260 Ethics-Safety Guidelines on July 1. Issued by the National Information Security Standardization Technical Committee, this document serves as the foundational ethical framework for AI applications in China.
While the TC260 guidelines are technically voluntary, they establish the normative standards that Chinese regulators will use to evaluate AI systems. The document outlines nine core ethical principles, emphasizing a “human-centric approach” in which AI must promote human well-being, respect life, and ensure “controllability and trustworthiness” while preserving human autonomy.
Accompanying the ethical guidelines are two new national standards (GB/T) focused on data security. These standards establish the technical baseline for personal information portability, allowing users to transfer their data between different AI services—and define the methodologies for personal information protection compliance audits.
A Global Regulatory Divergence
The implementation of these three frameworks highlights a growing divergence in global AI governance. While the European Union’s AI Act focuses heavily on classifying systems by risk tiers, and the United States relies largely on voluntary commitments and fragmented state-level laws, China is aggressively regulating specific use cases and interaction modalities.
By establishing clear, binding rules for autonomous agents and emotional AI, Beijing is attempting to front-run the next wave of technological disruption. The regulations demonstrate a sophisticated understanding of where AI is heading, aiming to harness the economic benefits of agentic systems while strictly controlling the societal impacts of increasingly human-like artificial intelligence.
