China Global AI Governance Plan

Overview

The Global AI Governance Action Plan was published on 26 July 2025 at the 2025 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance. China’s Permanent Mission to the United Nations publishes its English text. The plan presents artificial intelligence as both a development opportunity and a source of unprecedented risks. It advocates global solidarity to promote the safety, reliability, controllability, and fairness of AI development and use. It is a non-binding international policy agenda, rather than a statute, binding administrative regulation, or treaty. Its adoption represents the issuance of a policy, not legislative passage.

The plan connects technological development with economic inclusion, public services, sustainability, and international cooperation. Its principles include AI for good and service to humanity, respect for national sovereignty, alignment with development goals, safety and controllability, fairness and inclusiveness, and open cooperation. Its geographic scope is global cooperation advanced by China, rather than a domestic territorial regulatory regime. Its stakeholders include governments, public-sector bodies, international organizations, AI enterprises and other businesses, research institutions, academia, social organizations, individual citizens, and national and international standard-setting bodies.

Terminology and Policy Concepts

The plan describes artificial intelligence as a frontier of human development, a driver of scientific and industrial transformation, and an international public good benefiting humanity. These descriptions express the policy’s orientation toward shared development and access. References to AI systems, models, services, training data, and applications identify the technologies and activities addressed by the agenda. The plan’s governance proposals concern their development, deployment, safety, and social effects across different sectors and national development conditions.

“AI Plus” refers to exploring application scenarios and integrating AI across economic and social activities. Open-source ecosystems encompass communities, platforms, technical documentation, API documentation, and compatible upstream and downstream products. Categorized and tiered management is a proposed differentiated approach to AI governance. Interpretability, transparency, safety, and traceability are objectives for responsible development and use. The Global South is a priority beneficiary of infrastructure cooperation and capacity building. Digital and intelligent rights and interests feature in the plan’s attention to women and children during capacity-building activities.

Governance and Institutional Framework

The institutional approach is multilateral and multi-stakeholder. The plan calls for participation by governments, international organizations, industry, academia, social organizations, and individuals, supported by platforms for dialogue and cooperation. It identifies the United Nations as the main channel for implementing a fair and inclusive global digital governance system. This approach is linked to international law, national sovereignty, and differences in national development conditions. The proposed framework combines international coordination with respect for countries’ development choices and governance responsibilities.

The plan supports the establishment and early operation of the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance under the UN framework. It recognizes roles for the International Telecommunication Union, International Organization for Standardization, and International Electrotechnical Commission in standards cooperation. Public-sector bodies are encouraged to lead in reliable AI deployment and governance. The institutional agenda emphasizes dialogue, participation, scientific cooperation, and standards development rather than centralized international enforcement.

Development and Technical Priorities

The development agenda covers innovation, sectoral deployment, infrastructure, open ecosystems, data availability, and sustainability. The plan calls for international research cooperation, an innovation-friendly policy environment, policy and regulatory coordination, and reduced technology barriers. It promotes applications in manufacturing, consumption, commercial circulation, healthcare, education, agriculture, poverty alleviation, autonomous driving, and smart cities. Infrastructure priorities include clean power, next-generation networks, intelligent computing, and data centers. It also promotes infrastructure interoperability and a unified computing-power standards system.

Open-source cooperation is intended to lower innovation barriers, facilitate resource sharing, and support secure platforms and compatible products. The plan promotes sharing technical and API documentation and developing an open-source compliance system alongside technical safety guidance. Data priorities include high-quality datasets, lawful and orderly data flows, diverse corpora, and protection of cultural diversity. Sustainability measures include energy and water efficiency standards, low-power chips, efficient algorithms, and environmental applications of AI. Together, these priorities connect access to computing, data, and knowledge with resource-efficient innovation.

Implementation and Capacity Building

Implementation is organized around cooperation and shared capabilities. Proposed vehicles include international scientific platforms, cross-border application projects, open-source communities, shared infrastructure, data-sharing mechanisms, and exchanges of best practices. Capacity-building measures include joint laboratories, education and training, industry supply-demand matchmaking, and collaborative development of datasets and corpora. The plan also calls for mutual-recognition platforms for safety assessment. These mechanisms are intended to support both technological development and the ability to manage the risks associated with AI.

Developing countries are a central focus of implementation. The plan promotes their ability to access, use, innovate with, and govern AI through infrastructure support, training, research partnerships, and shared resources. It also seeks improved public AI literacy and attention to women’s and children’s digital and intelligent rights and interests. Organizations pursuing voluntary alignment can connect projects to these priorities, including secure development, public-service safety, documentation sharing, and capacity-building outcomes. Such alignment is distinct from compliance with independently applicable law governing data, procurement, intellectual property, or other activities.

Safety, Monitoring, and Evaluation

The plan proposes timely risk assessment, targeted prevention and response, categorized and tiered management, testing and evaluation, secure development, and traceability. These measures address AI safety risks and threats throughout development and deployment. It promotes information sharing and emergency-response development, alongside improvements in interpretability, transparency, and safety. Privacy, data security, and discrimination prevention are important governance priorities. Traceability management for AI services is intended to help prevent misuse and abuse.

Public-sector deployment receives specific attention. The plan calls for regular safety assessment of AI systems used in public services such as healthcare, education, and transportation. It also supports a broader risk testing and evaluation system and international safety-assessment cooperation. Organizations implementing these priorities voluntarily can document testing results, response exercises, infrastructure access, training participation, and cooperation outcomes as part of their own governance practices. These implementation choices should be distinguished from the policy’s recommendations, which do not constitute mandatory conformity assessment or reporting duties.

Legal Character and Accountability

The plan’s legal character is non-binding. Its calls for cooperation, safety assessment, secure development, and responsible deployment are policy recommendations rather than enforceable obligations under this instrument. It does not create a separate sanctions or appeals regime. The document’s references to privacy, data security, intellectual property, and responsible use guide cooperation and governance priorities without creating a standalone cause of action. Voluntary participation in a proposed cooperation mechanism is therefore distinct from satisfying a statutory compliance requirement.

The plan emphasizes lawful data flows, compliant data utilization, and respect for patents and software copyrights. Activities involving data collection, model development, public procurement, intellectual-property use, and international collaboration remain subject to independently applicable law. The plan is not an authorization for cross-border data transfers, an intellectual-property license, or an exemption from domestic requirements. Organizations using it to guide projects should distinguish its policy objectives from legal duties governing their own operations and from any contractual commitments made through separate cooperation arrangements.

Relationship to Other Instruments

The plan links its agenda to the United Nations Pact for the Future and its annex, the Global Digital Compact. It also supports implementation of the UN 2030 Agenda for Sustainable Development. These connections underpin its focus on digital inclusion, developing-country participation, sustainability, and shared governance. They situate the plan within a wider international development and digital-governance agenda while preserving its character as a policy proposal for cooperation.

The Global AI Governance Initiative, officially dated 20 October 2023, sets out related principles concerning sovereignty, people-centered development, risk-based governance, privacy, ethics, and developing-country representation. The Ministry of Foreign Affairs’ Artificial Intelligence Issues policy index also lists the Shanghai Declaration on Global AI Governance, dated 4 July 2024; the AI Capacity-Building Action Plan for Good and for All, dated 27 September 2024; and the AI+ International Cooperation Initiative, dated 24 September 2025. These instruments form part of China’s wider international AI policy activity.

International Alignment

International alignment is framed as cooperation that respects sovereignty and differences in development conditions. The plan promotes policy and regulatory coordination, dialogue among national standard-setting bodies, and participation by ITU, ISO, and IEC. It seeks inclusive and interoperable standards addressing security, industry, and ethics. Cross-border application partnerships, open-source communities, safety-assessment cooperation, and infrastructure support provide complementary channels for collaboration. The standards agenda aims to improve compatibility and shared understanding across AI ecosystems.

The plan combines technological capacity building with participation in governance. Developing countries are encouraged to contribute to discussions about AI development and safety, supported by infrastructure, joint laboratories, training, datasets, and improved public literacy. UN mechanisms are endorsed as avenues for meaningful dialogue and safe, fair, and inclusive development. The proposed global data-sharing platform is an exploratory cooperation measure. These proposals concern international coordination rather than a binding harmonized regulatory code or automatic equivalence between national AI rules.

Policy Timeline

Voluntary Alignment Checklist

Sources and References