China - AI Governance Principles (2019)

Overview

The "Governance Principles for a New Generation of Artificial Intelligence: Develop Responsible Artificial Intelligence" was released on 17 June 2019 by the National New Generation Artificial Intelligence Governance Expert Committee established under China’s Ministry of Science and Technology. It is a non‑binding policy framework that defines eight core principles to guide the research, development, deployment and governance of AI systems in China. The document situates AI governance within broader goals of sustainable economic, social and ecological development and calls for lifecycle governance, risk anticipation, and multi‑stakeholder cooperation. The English translations and media coverage are widely available; see the Ministry’s announcement and an English translation maintained by external research organizations for reference. Governance Principles (MOST announcement) and translation (New America / DigiChina). (most.gov.cn)

Definitions

The Principles do not offer a lengthy definitional section but rely on working meanings commonly used in policy discourse. "New generation artificial intelligence" is used to describe advanced AI systems characterized by cross‑domain integration, human‑machine collaboration, and collective intelligence. Key conceptual terms in the Principles include "safety and controllability" (meaning transparency, interpretability, robustness and traceability), "shared responsibility" (allocation of duties among developers, deployers, users and regulators), and "agile governance" (adaptive governance mechanisms that evolve with technology). The text assumes existing legal definitions of personal information, data controllers/processors and system beneficiaries where required by sectoral law, and delegates operational definitions to standards bodies and later implementing documents.

Governance and Institutional Framework

The Principles were drafted by the National New Generation Artificial Intelligence Governance Expert Committee (a MOST‑linked expert body) and published through official channels, positioning the committee as a central convenor of governance thinking. The release reinforces a governance architecture in China that combines central government leadership, multi‑agency coordination (e.g., Ministry of Science and Technology, Central Cyberspace Administration, Ministry of Industry and Information Technology), expert committees, standards organizations, and industry alliances. The document explicitly recommends that governance be collaborative — involving government, industry, research institutions, social organizations and the public — and that it be supported by technical standards, ethics norms, and normative instruments developed by standards bodies and sectoral authorities. For the official text and context, see the Ministry announcement: MOST – Governance Principles release. (most.gov.cn)

Key Focus Areas

The Principles are structured around eight focal areas that together map the intended outcomes and the governance levers: 1) Harmony and human‑friendliness: AI should enhance human welfare and respect human dignity; it should be developed to serve the progress of human civilization and societal well‑being. 2) Fairness and justice: The document calls for active efforts to eliminate bias and discrimination across data, algorithms, and application processes. 3) Inclusiveness and sharing: AI should support inclusive development, narrow regional disparities, reduce the digital divide, and avoid monopolistic data/platform consolidation. 4) Respect for privacy: The Principles demand boundaries and rules for collection, storage, processing and use of personal data, and mechanisms for consent and revocation of authorization. 5) Safety and controllability: Emphasis is placed on transparency, interpretability, reliability, robustness and traceability, and on the development of safety assessment and control capabilities. 6) Shared responsibility: Developers, deployers, users and other stakeholders should bear social responsibility and self‑discipline, and clear accountability mechanisms should be established. 7) Openness and collaboration: Cross‑disciplinary, cross‑boundary and international cooperation is encouraged; standards and shared norms are promoted to the extent practicable. 8) Agile governance: Governance must be adaptive, forward‑looking, and lifecycle oriented, allowing timely detection and mitigation of emergent risks. These focus areas are normative rather than prescriptive: the Principles outline policy aims and expectations rather than prescriptive compliance steps. Nevertheless, they provide the conceptual scaffolding for sectoral rules and standards that followed. (most.gov.cn)

Implementation Framework

The Principles envisage implementation by a combination of measures: voluntary adoption by industry; development of technical and ethical standards by standards bodies and expert groups; regulatory and administrative measures by relevant ministries and agencies; capacity building in safety testing and evaluation; and establishment of accountability and audit mechanisms. The text recommends integrating the Principles into the full lifecycle of AI products and services, promoting technical research on bias reduction, explainability, robustness and privacy‑protecting techniques, and encouraging industry self‑discipline pledges and codes of conduct. Implementation responsibility is distributed — the expert committee sets the high‑level direction, ministries and agencies adopt or reference the Principles in sectoral regulations or guidance, and standards organizations translate the guidelines into technical norms. For the original release and policy context see the government announcement. Chinese Government (gov.cn) announcement. (gov.cn)

Monitoring and Evaluation

The document advocates continuous monitoring, safety assessments, and iterative governance: agencies and expert groups should establish mechanisms for periodic review, risk scanning, safety verification and post‑market monitoring where applicable. It encourages the creation of audit trails, traceability features and recordkeeping to enable oversight and investigation. While the Principles do not establish a single supervising authority or a formal audit regime, they call for multi‑stakeholder monitoring networks and standards for testing and evaluation to be developed by technical and standards bodies in coordination with regulators and industry. External research organizations and think tanks have used the Principles as anchors when tracking Chinese AI governance evolution.

Penalties, Liability, and Appeals

Because the Principles are non‑binding guidance, they do not specify penalties or a direct administrative enforcement scheme. Instead, the text emphasizes shared responsibility and accountability mechanisms and encourages the creation of norms that will later be operationalized by binding legislation, sectoral administrative measures, or standards. In practice, enforcement and penalties for AI‑related harms in China are handled through existing legal frameworks (for example, cybersecurity, data protection and consumer protection regimes) or by later issued administrative rules and sectoral measures that implement specific obligations. Stakeholders should therefore treat the Principles as policy direction that will inform binding measures and potential liability allocations in subsequent regulatory instruments. (most.gov.cn)

Relationship to Other Instruments

The 2019 Governance Principles sit within a broader Chinese AI policy ecosystem. They follow the State Council’s 2017 "New Generation Artificial Intelligence Development Plan" and pre‑date later ethics and norm documents such as the 2021 "Ethical Norms for New Generation Artificial Intelligence". They also interact with data protection instruments, most notably the Personal Information Protection Law (PIPL) and the Cybersecurity Law, which provide binding rules for data handling, privacy, and network security. Standards bodies, industry alliances and technical committees have used the Principles as a normative reference when drafting technical standards and best practices. The Principles therefore act as a bridge between high‑level strategy and later, operational legislation and standards.

International Alignment

The Principles explicitly encourage international dialogue and cooperation, and they frame China’s AI governance objectives in terms of global public goods and a "community of common destiny for humanity." The document acknowledges that different countries may adopt different governance practices but promotes formation of international consensus on frameworks, standards and norms. Since 2019, China has used the Principles as part of its diplomatic and multilateral engagements on AI governance and standards development, and the text has been cited in comparative analyses of global AI ethics and governance frameworks. China Daily summary (English).

Implementation Timeline

Sources and References