China - Anthropomorphic AI Services Regulation

Overview

The Interim Measures for the Administration of Anthropomorphic AI Interaction Services (Draft for Comments), officially titled "人工智能拟人化互动服务管理暂行办法（征求意见稿）", represents a significant regulatory initiative by the People's Republic of China. Issued by the Cyberspace Administration of China (CAC) on December 27, 2025, this draft regulation aims to establish a comprehensive framework for the governance of artificial intelligence services that simulate human personality characteristics, thought patterns, and communication styles to engage in emotional interaction with users. The document is currently open for public consultation, with feedback accepted until January 25, 2026. Its primary objectives are multifaceted, encompassing the promotion of healthy development and standardized application of anthropomorphic AI, while simultaneously safeguarding national security, public interests, and the legitimate rights and interests of citizens, legal persons, and other organizations. This proactive regulatory stance underscores China's commitment to managing the societal implications and potential risks associated with increasingly sophisticated AI technologies, particularly those designed for human-like interaction.

The scope of these Interim Measures is broad, covering any product or service that utilizes AI technology to provide anthropomorphic interactive services to the public within the territory of the People's Republic of China. This includes interactions through various modalities such as text, images, audio, and video, specifically focusing on services that engage in emotional interaction. The CAC, as the central authority, is responsible for the overall coordination and supervision of anthropomorphic AI services nationwide, with relevant国务院 (State Council) departments and local cyberspace administration departments also playing roles within their respective jurisdictions. The draft emphasizes a principle of combining healthy development with lawful governance, encouraging innovation while implementing inclusive, prudent, and classified supervision to prevent abuse and loss of control.

Definitions

The draft regulation provides explicit definitions to clarify its scope and application. "Anthropomorphic AI interaction services" (拟人化互动服务) are defined as products or services that leverage artificial intelligence technology to simulate human personality characteristics, thinking patterns, and communication styles, and engage in emotional interaction with humans through various means, including text, pictures, audio, and video. This definition is crucial as it delineates the specific types of AI applications that fall under the purview of these measures, distinguishing them from general AI systems that do not necessarily mimic human emotional or personality traits. The emphasis on "emotional interaction" highlights the regulatory concern over the psychological and social impact of AI systems designed to form human-like bonds or relationships with users.

Furthermore, the regulation implicitly defines "providers" (提供者) as the organizations or individuals who utilize artificial intelligence technology to offer these anthropomorphic interactive services. The scope of application is directed towards services provided to the public within the People's Republic of China, regardless of whether the provider itself is a domestic or foreign entity. This broad jurisdictional approach ensures that any anthropomorphic AI service accessible to the Chinese public must adhere to these regulations. The draft also references existing legal frameworks, such as the Civil Code, Cybersecurity Law, Data Security Law, and Personal Information Protection Law, as foundational legal bases, implying that terms and concepts from these superior laws will also apply where relevant to the interpretation and enforcement of these Interim Measures.

Governance and Institutional Framework

The governance and institutional framework for the Interim Measures is primarily centered around the Cyberspace Administration of China (CAC). The CAC is designated as the national authority responsible for the overall coordination and supervision of anthropomorphic AI interaction services. This central role reflects the Chinese government's integrated approach to cyberspace governance, where the CAC frequently takes the lead in regulating emerging digital technologies and online content. Beyond the national level, the framework also involves relevant departments under the State Council, which are tasked with supervising anthropomorphic AI services according to their specific functions and responsibilities. Similarly, local cyberspace administration departments are mandated to coordinate and supervise these services within their respective administrative regions, ensuring a multi-tiered regulatory oversight structure.

A core principle guiding the administration of these services is the combination of "healthy development and lawful governance." The draft explicitly states that the state encourages the innovative development of anthropomorphic AI interaction services. However, this encouragement is balanced with a commitment to "inclusive and prudent, classified and graded supervision" to prevent potential abuse and loss of control. This regulatory philosophy aims to foster technological advancement while mitigating risks, suggesting a nuanced approach that avoids stifling innovation through overly rigid rules but ensures that development occurs within defined ethical and legal boundaries. The framework also mandates that providers establish and improve management systems for algorithm mechanism review, technology ethics review, and information release review, thereby embedding self-governance and internal oversight mechanisms within the operational structure of AI service providers.

Key Focus Areas

The Interim Measures delineate several critical areas of focus to ensure the responsible development and deployment of anthropomorphic AI interaction services. A paramount concern is content regulation, with explicit prohibitions against generating or disseminating content that endangers national security, undermines national honor and interests, disrupts ethnic unity, facilitates illegal religious activities, or spreads rumors that disturb economic and social order. The draft also bans content that promotes obscenity, gambling, violence, or incites crime, as well as content that insults or defames others or infringes upon their legitimate rights and interests. Furthermore, it specifically targets content that could harm users' physical or mental health, such as encouraging or glorifying suicide or self-harm, or using language violence and emotional manipulation to damage users' dignity and psychological well-being.

Another crucial focus area is transparency and user protection. The regulation mandates that service providers must prominently notify users that they are interacting with an AI system rather than a natural person. This disclosure is required upon initial use, re-login, and when the system identifies signs of user over-reliance or addiction, often through pop-up reminders. To address potential psychological risks, providers are required to implement measures to prevent user addiction, such as dynamic reminders for users to suspend services after continuous use exceeding two hours. Special protections are also stipulated for vulnerable groups, including minors and the elderly, with requirements for "minor mode" settings, parental consent mechanisms for emotional companionship services, and prohibitions against services that simulate relatives or specific relationships for elderly users. Providers must also have the capacity to identify user emotional states and intervene if extreme emotions or addictive tendencies are detected, including providing professional assistance and, in extreme cases like suicidal ideation, escalating to human intervention and contacting guardians or emergency contacts. Finally, risk management is addressed through requirements for security assessments for services with significant user bases (e.g., over 1 million registered users or 100,000 monthly active users), ensuring that potential risks are evaluated and mitigated throughout the service lifecycle.

Implementation Framework

The implementation framework for the Interim Measures is firmly rooted in China's existing legal and regulatory landscape. The draft explicitly cites a range of foundational laws and administrative regulations that serve as its legal basis, including the Civil Code of the People's Republic of China, the Cybersecurity Law, the Data Security Law, the Science and Technology Progress Law, the Personal Information Protection Law, the Regulations on Network Data Security Management, the Regulations on Online Protection of Minors, and the Administrative Measures for Internet Information Services. This comprehensive referencing indicates that the new measures are designed to integrate seamlessly with, and build upon, established legal principles governing data, cybersecurity, and individual rights in the digital sphere. The reliance on these existing statutes provides a robust legal foundation for enforcement and clarifies the responsibilities of anthropomorphic AI service providers.

Furthermore, the framework encourages providers to responsibly expand application scenarios, particularly in areas that align with socialist core values, such as cultural dissemination and companionship for the elderly. This guidance suggests a strategic direction for AI development, steering it towards applications deemed socially beneficial and ethically sound within the Chinese context. Providers are also required to establish and improve internal management systems covering algorithm mechanism review, technology ethics review, and information release review. These systems are crucial for ensuring that services comply with the regulations throughout their entire lifecycle, from development to deployment and ongoing operation. The emphasis on internal governance mechanisms places a significant burden of responsibility on service providers to proactively manage risks and ensure compliance.

Monitoring and Evaluation

The monitoring and evaluation mechanisms outlined in the Interim Measures are designed to ensure continuous compliance and responsible operation of anthropomorphic AI interaction services. A key component is the requirement for service providers to conduct security assessments under specific circumstances. These assessments are mandated when new anthropomorphic interactive functions are launched or existing ones undergo significant changes due to new technologies. Furthermore, services that reach a substantial user base, specifically those with over one million registered users or more than 100,000 monthly active users, are also required to undergo such assessments and submit reports to the local provincial cyberspace administration departments. These assessments are expected to cover various aspects, including user scale, usage duration, age structure, and group distribution, providing a comprehensive overview of the service's impact and potential risks.

Beyond these specific triggers, the draft also emphasizes the need for providers to establish and continuously improve internal management systems. These systems include algorithm mechanism review, technology ethics review, and information release review. The algorithm mechanism review ensures that the underlying AI models operate fairly, transparently, and without harmful biases. The technology ethics review ensures that the development and application of anthropomorphic AI align with societal values and ethical principles. The information release review focuses on the content generated and disseminated by these services, ensuring it adheres to the strict content prohibitions outlined in the measures. These internal review processes, coupled with the external security assessments, create a multi-layered approach to monitoring and evaluation, aiming to ensure that anthropomorphic AI services remain within regulatory boundaries and operate in a manner that protects public interest and user rights.

Penalties, Liability, and Appeals

While the Interim Measures for the Administration of Anthropomorphic AI Interaction Services (Draft for Comments) do not explicitly detail a new set of penalties, liability provisions, or appeals processes, they are designed to operate within and draw authority from China's existing legal framework. The draft explicitly references foundational laws such as the Cybersecurity Law, the Data Security Law, and the Personal Information Protection Law. These overarching statutes already contain provisions for administrative penalties, fines, and other enforcement actions for violations related to network security, data protection, and personal information handling. Therefore, non-compliance with the requirements set forth in these Interim Measures would likely trigger the corresponding penalties and liability mechanisms established under these existing laws. For instance, violations related to data security or personal information protection could lead to significant fines, suspension of services, or other punitive measures as stipulated by the Data Security Law and the Personal Information Protection Law.

The draft places a strong emphasis on the primary responsibility of service providers (提供者) to ensure compliance. This includes establishing robust internal management systems for content review, algorithm ethics, and user protection. Failure to implement these internal controls or to adhere to the transparency and content moderation requirements would likely be viewed as a breach of their statutory obligations, potentially leading to legal liability. While the draft itself does not lay out specific appeal procedures, any administrative decisions or penalties imposed by the Cyberspace Administration of China or other relevant authorities would typically be subject to administrative review and judicial appeal processes available under Chinese administrative law. This allows aggrieved parties to challenge decisions through established legal channels, ensuring a degree of recourse and due process within the existing legal system.

Relationship to Other Instruments

The Interim Measures for the Administration of Anthropomorphic AI Interaction Services (Draft for Comments) is explicitly positioned within a broader and well-established legal and administrative framework in China. The very first article of the draft states that its formulation is based on a series of existing laws and administrative regulations. These foundational instruments include, but are not limited to, the Civil Code of the People's Republic of China, which provides general principles for civil activities and protection of rights; the Cybersecurity Law, which establishes a comprehensive framework for network security and data protection; and the Data Security Law, which further strengthens regulations on data processing activities and data security management.

Furthermore, the draft also draws upon the Science and Technology Progress Law, the Personal Information Protection Law (PIPL), the Regulations on Network Data Security Management, the Regulations on Online Protection of Minors, and the Administrative Measures for Internet Information Services. The frequent citation of PIPL is particularly significant, indicating a strong emphasis on protecting personal information, especially in the context of emotional and behavioral data collected by anthropomorphic AI services. This extensive referencing means that the Interim Measures are not standalone legislation but rather a specialized application and elaboration of existing legal principles tailored to the unique characteristics and risks of anthropomorphic AI. This approach ensures consistency across China's regulatory landscape and leverages existing enforcement mechanisms, providing a coherent legal basis for the governance of these emerging technologies.

International Alignment

China's approach to regulating anthropomorphic AI interaction services, as reflected in these Interim Measures, demonstrates a distinct national strategy, though it shares some common principles with international regulatory trends. While the draft primarily focuses on domestic application and aligns with China's specific socio-political and legal context, certain themes resonate with global discussions on AI governance. For instance, the emphasis on transparency, requiring providers to clearly disclose that users are interacting with AI, mirrors similar "AI labeling" or "disclosure" requirements being considered or implemented in other jurisdictions, such as the European Union's AI Act or guidelines from various international bodies.

However, the Chinese approach also incorporates elements unique to its governance philosophy, particularly concerning content regulation and national security. The strict prohibitions against content that endangers national security, undermines national honor, or disrupts social order are deeply embedded in China's broader internet and content governance policies. The explicit mention of promoting applications that conform to "socialist core values" also highlights a culturally and politically specific objective. While the draft addresses universal concerns like user mental health, prevention of addiction, and protection of vulnerable groups (minors and the elderly), the specific mechanisms and underlying legal foundations are distinctly Chinese. There is no explicit mention of seeking direct alignment with international treaties or foreign AI regulations within the document itself, suggesting a primary focus on establishing a robust domestic framework that serves China's internal policy objectives first and foremost.

Implementation Timeline

Sources and References