China AI Ethics and Safety Guidelines

Overview

The Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0, issued by China's National Technical Committee 260 on Cybersecurity (TC260) on May 19, 2026, represent a landmark effort to provide comprehensive guidance on ethical and safety considerations across the entire lifecycle of AI applications in China. This document is a principle-based and reference technical guideline, emphasizing a human-centered approach to AI development, ensuring that AI serves the common good, and maintaining human leadership over AI systems. It addresses the rapidly evolving landscape of AI, including emerging innovations such as intelligent agents and integrated AI systems, which are becoming increasingly autonomous, complex, and influential. The guidelines aim to balance technological advancement with the mitigation of ethical and safety risks, such as job displacement, decision-making responsibility, and potential emotional dependence on AI, thereby safeguarding human agency, social trust, and public order.

The release of these guidelines underscores China's strategic commitment to fostering responsible AI innovation while mitigating potential societal and security risks. It reflects a broader national strategy to position China as a global leader in AI by 2030, integrating AI development with national strategic goals and social stability. The guidelines are designed to provide practical direction for organizations and individuals involved in AI-related activities, from development and service provision to end-use. They articulate a set of core ethics-safety principles intended to steer the design, deployment, and operation of AI systems in a manner that promotes human well-being, fairness, transparency, and control, while also establishing mechanisms for risk assessment, accountability, and continuous improvement.

Definitions

The Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0 establish a foundational glossary of terms crucial for understanding and implementing its provisions. Key definitions include "Artificial Intelligence (AI) System," referring to intelligent systems capable of autonomous perception, memory, decision-making, interaction, and execution, rapidly integrating with cyberspace and the physical world, especially with the emergence of advanced technologies like large language models. "Ethical AI" is defined as AI systems designed and deployed in alignment with human values, promoting fairness, transparency, accountability, and respect for fundamental rights. "AI Safety" encompasses the robustness, reliability, security, and controllability of AI systems, ensuring they operate as intended without causing unintended harm or risks to individuals, society, or national security.

Further definitions include "Risk Assessment," which refers to the systematic process of identifying, analyzing, and evaluating potential ethical and safety risks associated with AI applications throughout their lifecycle. "Human Oversight" is emphasized as the imperative for human judgment, supervision, intervention, and correction at key stages of AI system operation, preventing excessive replacement of human decision-making. "Developers" are entities responsible for the design, training, and testing of AI systems, while "Service Providers" are those deploying and offering AI-powered products or services to the public. "Controllability and Trustworthiness" are defined as the ability to understand, predict, and manage the behavior of AI systems, ensuring their reliability and adherence to specified performance and ethical benchmarks. These definitions provide a common language and framework for all stakeholders engaged in AI development and deployment within China.

Governance and Institutional Framework

The governance and institutional framework for the Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0 is primarily anchored by the National Technical Committee 260 on Cybersecurity (TC260), which is China's key body responsible for cybersecurity standardization and the issuer of these guidelines. TC260 plays a pivotal role in developing national standards and technical specifications that guide the ethical and safe development and deployment of AI. Beyond TC260, a multi-agency approach characterizes China's AI governance, involving several other significant governmental bodies. The Cyberspace Administration of China (CAC) is identified as the primary and lead authority for regulating AI, responsible for rulemaking, enforcement, and coordinating AI governance across departments.

Other crucial entities include the Ministry of Industry and Information Technology (MIIT), which focuses on AI industry development, technical standards establishment, and industrial applications. The Ministry of Science and Technology (MOST) oversees national AI ethical guidelines and research. The National Development and Reform Commission (NDRC) also plays a role in strategic planning and promoting standardized application and innovative development of AI. The State Administration for Market Regulation (SAMR) handles national standardization related to AI products. This collaborative framework ensures a comprehensive oversight mechanism, integrating ethical considerations with technological advancement, industrial policy, and national security objectives, reflecting China's emphasis on state-led initiatives and coordinated governance in the AI sector.

Key Focus Areas

The Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0 delineate nine core ethics-safety principles that form the bedrock of responsible AI development and deployment in China. These principles include: enhancing human welfare, respecting the right to life, fairness and justice, reasonable risk control, openness and transparency, privacy and security, controllability and trustworthiness, agile co-governance, and inclusive benefit-sharing. Each principle is designed to address specific ethical and safety challenges posed by AI. For instance, "enhancing human welfare" emphasizes promoting AI for human development and the common good, while also protecting employment rights and assessing potential job displacement. "Fairness and justice" aims to ensure equal opportunity and eliminate bias and discrimination in AI systems.

Furthermore, the guidelines identify six major ethical risks that stakeholders must actively mitigate: weakening of human oversight, disruption of public order, misinformation and knowledge distortion, algorithmic discrimination, infringement of individual rights, and damage to ecosystems. To address these, "reasonable risk control" mandates the integration of risk assessment frameworks and mitigation strategies throughout the AI lifecycle. "Openness and transparency" requires traceable logs and explainability reports for AI systems, while "privacy and security" necessitates encrypted data storage and robust data protection measures. "Controllability and trustworthiness" focuses on ensuring that humans retain real decision-making power and can intervene or terminate AI operations at any time. These principles and risk areas provide a comprehensive framework for guiding AI development towards beneficial and safe outcomes, aligning with China's broader goals of technological progress and social stability.

Implementation Framework

The implementation framework for the Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0 provides operational guidance tailored for various stakeholders, including AI developers, service providers, and users. Developers are mandated to integrate ethics-safety considerations throughout the entire AI development process, from data collection and model training to testing and deployment. This includes meeting China's AI ethics review requirements, acknowledging the inherent capability boundaries of AI, designing systems that are inclusive and consider vulnerable groups, enhancing transparency in their models, and establishing robust incident traceability mechanisms. Developers are also expected to embed risk controls proactively during the data collection, training, and testing phases to ensure the intrinsic safety and ethical alignment of AI systems from their inception.

Service providers, who deploy and offer AI applications to the public, are tasked with selecting ethical AI objectives and ensuring that AI consistently serves an auxiliary role in critical decision-making processes, particularly in high-risk scenarios such as finance, healthcare, education, and areas impacting national security or public interest. They must provide clear mechanisms for user intervention, minimize the processing of sensitive personal information, and adequately notify users about data collection practices. Additionally, service providers are required to establish secure deployment environments, implement real-time monitoring, and develop comprehensive emergency response mechanisms. Users, in turn, are advised to exercise moderate AI use, understand the nature of AI emotional services, and avoid excessive dependence or replacing real interpersonal communication with AI, emphasizing AI's role as a tool to assist real life rather than dictate it.

Monitoring and Evaluation

Monitoring and evaluation under the Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0 are designed to ensure continuous adherence to ethical and safety principles and to assess the effectiveness of implemented measures. The framework emphasizes the establishment of robust systems for oversight and accountability across the AI lifecycle. Key operational duties detailed in the guidelines include the requirement for every party involved in AI activities—developers, providers, and users—to retain audit logs for a specified period, typically ten years. This comprehensive logging is crucial for satisfying government oversight demands and enabling thorough post-incident analysis and accountability. Continuous model review cycles are also established as a core operational rhythm, ensuring that AI systems are regularly assessed for performance, bias, and adherence to ethical standards.

To strengthen the cybersecurity posture and ethical compliance, the guidelines promote periodic bias audits, privacy impact assessments, and penetration tests. These assessments help identify and mitigate potential vulnerabilities and discriminatory outcomes. Furthermore, the establishment of multidisciplinary safety boards, encompassing legal, engineering, and ethics experts, is encouraged to provide holistic oversight and guidance. Incident response plans are mandated to include rapid reporting windows, typically within 24 hours, for critical security or ethical breaches. For critical systems, human approval loops are indispensable, ensuring meaningful human control and intervention. The guidelines also anticipate future drafts from TC260 that will sharpen metrics for transparency and fairness, indicating an evolving and adaptive approach to monitoring and evaluation, aiming to build resilient AI programs that can innovate without regulatory shocks.

Penalties, Liability, and Appeals

As the Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0 are primarily a principle-based and reference technical document, they do not explicitly detail direct penalties for non-compliance. However, non-adherence to these guidelines can have significant implications for AI developers and service providers, particularly in relation to China's broader and increasingly stringent AI regulatory landscape. Non-compliance may lead to reputational damage, loss of public trust, and could impede an entity's ability to participate in state-backed AI initiatives or secure necessary licenses for AI-related services. More importantly, while the guidelines themselves are non-binding, their principles are often reflected in existing and forthcoming binding laws and regulations, meaning that violations of the guidelines could indirectly lead to penalties under other applicable laws.

For instance, the principles of data protection, cybersecurity, and algorithmic fairness articulated in these guidelines are strongly aligned with mandatory laws such as the Cybersecurity Law, the Data Security Law, and the Personal Information Protection Law (PIPL), as well as specific AI regulations like the Interim Measures for the Management of Generative AI Services. Violations of these binding laws, which incorporate the spirit of ethical and safety guidelines, can result in substantial fines, suspension of services, or even criminal liability for severe breaches. The guidelines encourage the establishment of internal accountability mechanisms and clear lines of responsibility for AI systems. While specific appeal processes for guideline violations are not outlined, stakeholders can generally seek redress through existing administrative review channels or judicial processes for violations of underlying binding laws. The emphasis is on proactive risk management and compliance to avoid triggering more severe legal consequences under the comprehensive Chinese regulatory framework.

Relationship to Other Instruments

The Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0 are designed to integrate seamlessly within China's evolving and comprehensive AI regulatory ecosystem, serving as a foundational ethical and safety framework that complements existing and forthcoming legal instruments. These guidelines explicitly supplement the 2025 AI Safety Governance Framework 2.0, providing more detailed ethical and safety considerations. They also build upon and are intended to be read in conjunction with China's foundational cybersecurity and data protection laws, including the Cybersecurity Law, the Data Security Law, and the Personal Information Protection Law (PIPL). These basic laws provide the legal backbone for data security, personal information protection, and algorithmic governance, which are critical components of ethical and safe AI.

Furthermore, the guidelines interact with a growing number of specific AI regulations that have been introduced in recent years. These include the Interim Measures for the Management of Generative Artificial Intelligence Services, the Administrative Provisions on Deep Synthesis in Internet-based Information Services, and the Administrative Provisions on Recommendation Algorithms in Internet-based Information Services. The principles outlined in the Ethics-Safety Guidelines inform the implementation and interpretation of these more specific regulations, ensuring a consistent ethical and safety posture across different AI applications. For example, the emphasis on transparency, fairness, and human oversight in the guidelines directly supports the requirements for content moderation, algorithm filing, and security assessments mandated by the generative AI and deep synthesis regulations. The guidelines also align with broader national strategies like the New Generation Artificial Intelligence Development Plan, which sets the overarching strategic objective for China's AI leadership by 2030.

International Alignment

While Chinese legislation, including these guidelines, does not directly reference international AI standards or guidelines, the principles and spirit embedded within the Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0 often echo those found in international AI regulatory frameworks. China's approach to AI governance, as reflected in these guidelines, emphasizes human-centered development, AI for good, and maintaining human control, which resonates with global discussions around responsible AI, such as the OECD AI Principles and UNESCO's Recommendation on the Ethics of Artificial Intelligence. The focus on principles like fairness, transparency, accountability, privacy, and security demonstrates a shared understanding of fundamental ethical considerations in AI development and deployment.

China's strategic goal of becoming a global leader in AI by 2030 inherently drives a need for its regulatory frameworks to be robust and credible on an international stage. Although the specific mechanisms and enforcement might differ due to distinct legal and political systems, the underlying objectives of mitigating algorithmic discrimination, ensuring data protection, and promoting safe and controllable AI systems align with international best practices. The guidelines contribute to the global discourse on AI ethics and safety by offering a Chinese perspective that integrates national strategic priorities with universal ethical concerns, potentially fostering future cross-border cooperation and mutual recognition in AI governance, especially as China actively participates in international standardization efforts for AI.

Implementation Timeline

Compliance Checklist

Sources and References