China on Tuesday released technical guidelines on the ethics and safety of artificial intelligence applications, proposing China’s approach that emphasizes human-centered development, AI for good, and human leadership for AI systems.
The document, titled “Ethical Safety Guidelines for Artificial Intelligence Applications 1.0”, was presented at the sub-forum of the 2026 China Internet Civilization Conference. The annual conference will be held in Nanning, the capital of Guangxi Zhuang Autonomous Region.
Published by the National Technical Commission on Cybersecurity 260, China’s main body responsible for cybersecurity standardization, the guidelines are described as a “principles-based reference technical document.” These guidelines cover the development, service provision, and use of AI applications and provide guidance to organizations and individuals engaged in AI-related activities.
It outlines key principles such as rationally managing risk, maintaining openness and transparency, protecting privacy and security, ensuring control and trust, agile co-governance, and inclusive sharing.
The guidelines say AI applications should always serve the “common well-being of humanity” and call for human control, emergency response and intervention mechanisms to be established at key stages to ensure “human control over AI applications remains”.
The document also urges stakeholders to maintain “human judgment, oversight, intervention, and correction” at key stages to avoid excessive replacement of human decision-making by AI.
Users are encouraged to use AI in moderation and to have a proper understanding of AI emotional services. It states that AI should be treated as a “tool that supports real life,” and calls for avoiding excessive dependence or dependence on it, as well as replacing actual interpersonal communication and activities with AI.
For information and communication services, the guidelines call for stronger risk governance of AI-generated and recommended content, warning against “information cocooning, cognitive misinterpretation, and cognitive decline.”
The guidelines also encourage the creation of open source innovation ecosystems, including open access to AI models, tool components, and evaluation benchmarks, while improving the security capabilities of open source ecosystems.
