Research Report on Global AI Governance

Research Report on Global AI Governance

 Jointly Launched by 

Shanghai Academy of Social Sciences

Wuhan University

Tongji University

China Institute of Contemporary International Relations

China Academy of Information and Communications Technology

Institute of World Economics and Politics of Chinese Academy of Social Sciences

Beijing University of Posts and Telecommunications

Beihang University

Beijing Institute of Technology

People's Public Security University of China

China University of Political science and Law

Fudan University

Nanjing University of Posts And Telecommunications

Zhejiang Normal University

Fuxi Institution

Preface

    Artificial intelligence (AI) technology represents the most emblematic disruptive force in the new round of technological revolution and industrial transformation. While presenting tremendous development opportunities for human society, it also poses many global challenges. The widespread adoption of AI technology has served the development of a wide range of sectors, including healthcare, education, transportation, agriculture, industry, finance, culture, and ecology. At the same time however, misuse, abuse, and ill-intended use of AI have jeopardized the security of individuals, groups, society, the eco-system, as well as value and norms in various dimensions. A new wave of global AI governance triggered by the breakthroughs made in generative AI technology in 2022 has been sweeping across the world .

    Against this backdrop, experts from the Shanghai Academy of Social Sciences, Wuhan University, Tongji University, China Institutes of Contemporary International Relations, China Academy of Information and Communications Technology, Institute of World Economics and Politics of Chinese Academy of Social Sciences, Beijing University of Posts and Telecommunications, Beihang University, Beijing Institute of Technology, People's Public Security University of China, China University of Political Science and Law and other think tanks and universities in China have conducted research on the basic status, key issues, and system building of global AI governance. The term "global AI governance" in this report refers to the collaborative process through which international organizations, governments, technology companies, non-governmental organizations (NGOs), and other actors jointly formulate and implement a series of principles, norms, standards, policies, laws, and institutions to ensure global security and peaceful use of AI. It needs to be pointed out that the application of AI technology in the military domain falls into a separate international regulatory process, and this research report does not cover the global governance process concerning the development or use of AI for military purposes.

Contents

Preface................................................................................................ I

I. Current Status of Global AI Governance................................... 1

1. Progress  in Developing Multilateral Platforms................... 1

2.  Achievements Made by Multi-party Mechanisms.............. 3

3. Practices  of Relevant Countries/Regions............................ 4

4. Operations  of Technology Companies................................. 5

II. Ten Key Issues in Global AI Governance................................. 7

1. Principle  of National Sovereignty and AI Development... 7

2. Social  Change and Sustainable Development..................... 8

3. Technological  Innovation and Industrial Development..... 8

4. Human-Machine  Emotions and Bioethics........................... 9

5. Content  Security Risks.......................................................... 9

6. Model  Algorithm Security Risks........................................ 10

7. Data  Security and Privacy Protection................................ 10

8. Product  Liability and Risks................................................. 11

9. Intellectual  Property Protection.......................................... 11

10. AI  Divide and International Collaboration...................... 12

III. Building a Sound Global AI Governance System................ 12

1.  Objectives and Purposes..................................................... 13

2.  Principles and Consensus.................................................... 13

2.1  Respect for National Sovereignty.............................. 13

2.2  Balancing Development and Security....................... 13

2.3  Adhering to Equality, Mutual Benefit, and Inclusiveness    14

3. Action  Taking Pathways...................................................... 14

3.1  People-Centered Approach......................................... 14

3.2.  Commitment to Intelligence for Good..................... 15

3.3  Empowering Various Industries................................. 15

3.4  Preventing and Managing Security Risks................. 15

3.5  International Cooperation Capacity Building........... 16

3.6  Improving Global Governance Mechanisms............ 16

Appendix: Overview of Documents on Global AI Governance  in

 Recent Years................................................................ 19

I. Current Status of Global AI Governance

    In 1956, the Dartmouth Conference officially established the term "Artificial Intelligence," marking the beginning of independent research in AI science. Since then, AI development has experienced ups and downs, and achieved some breakthroughs in the 1990s. In 1997, the Deep Blue supercomputer defeated the world champion in chess, sparking public contemplation about the future of technology and prompting early international discussions on AI ethics.

    As AI technology evolved towards machine learning, reinforcement learning, and deep learning, the landmark event of AlphaGo defeating the world champion in Go in 2016 ushered global AI governance into an exploratory phase of formulating principles and guidelines. For instance, in 2016, the Institute of Electrical and Electronics Engineers (IEEE) proposed the Ethically Aligned Design to guide the design and development of autonomous and intelligent systems. In 2017, industry leaders worldwide initiated the Asilomar AI Principles, outlining guidelines for technology development that is "beneficial to humanity."

    Since 2022, the advent and global popularity of ChatGPT have accelerated the transition of AI from "small models + discriminative approaches" to "large models + generative approaches," accompanied by a surge in voices and actions for strengthening global AI governance.

    1. Progress in Developing Multilateral Platforms

    Multilateral governance processes serve as the primary channels for global AI governance and are crucial for fostering universal consensus among the international community in AI-related fields.

    The United Nations (UN) has played a major leading role in global AI governance. In 2023, UN Secretary-General António Guterres announced the establishment of the High-Level Advisory Board on Artificial Intelligence, which released the final report Governing AI for Humanity in 2024, proposing action plans to enhance global cooperation. In 2024, the UN General Assembly adopted two resolutions, namely Enhancing International Cooperation in Capacity-building of Artificial Intelligence and Seizing the Opportunities of Safe, Secure and Trustworthy Artificial Intelligence Systems for Sustainable Development, significantly promoting international cooperation in AI. The Global Digital Compact adopted at the UN Future Summit also aims to "strengthen international governance of AI for the benefit of humanity." It authorizes the United Nations to establish mechanisms such as an International Scientific Panel on Artificial Intelligence and a Global Dialogue on AI Governance.

    Several UN agencies have also made progress in this field. UNESCO's ethical framework has garnered active participation from over 50 countries, while the International Telecommunication Union's annual Global Summit on AI for Good, held for seven consecutive years, has become a vital platform for global AI dialogue and exchange.

    Intergovernmental organizations have also included global AI governance on their agendas, and achieved corresponding outcomes. The G20 reiterated its commitment to "put people at the center and harness AI for the benefit and service of all humanity" at the 2023 New Delhi Summit. In 2022, the BRICS countries agreed to establish an artificial intelligence research group, aiming to strengthen practical cooperation in areas such as AI technology research and development, standard setting, and industrial applications. Organizations such as The OECD have also updated relevant recommendations.

    2. Achievements Made by Multi-party Mechanisms

    Multi-party mechanisms have actively promoted the development of responsible AI development and application in areas such as technical standards and ethical norms.

    Technological communities such as IEEE, the International Organization for Standardization (ISO), and the International Electrotechnical Commission (IEC) have published cutting-edge technical standards for AI, ethical guidelines for AI design, and management guides for AI systems. In 2023, IEEE launched a program to make AI ethics and governance standards freely accessible. Information technology - Artificial intelligence - Management systems (ISO/IEC 42001 standard) provides a framework for organizations to manage AI.

    Furthermore, some international organizations, conferences and forums have accelerated their actions in this field. In 2023, the World Economic Forum established the AI Governance Alliance, proposing recommendations for responsible development, open innovation, and international cooperation. The Working Group on Artificial Intelligence of the World Internet Conference released the Consensus on Developing Responsible Generative Artificial Intelligence which is dedicated to promoting responsible AI development. In 2024, the Global System for Mobile Communications Association launched the Responsible AI Maturity Roadmap, which helps telecommunications companies assess their level of maturity in responsible AI.

    3. Practices of Relevant Countries/Regions

    Relevant countries and regions have proposed governance solutions with unique characteristics based on factors such as political culture, industrial development, and goals and aspirations.

    Upholding the philosophy and purpose of "being people-centered" and "intelligence for good" put forward by President Xi Jinping, China has made good progress in coordinating AI development and security. In 2023, China put forward the Global Initiative on AI Governance, detailing the country’s approach to AI governance in terms of its development, security, and governance. The Cyberspace Administration of China and six other departments jointly issued the Interim Measures for the Administration of Generative AI Services which is the world’s first legislation on generative AI. In 2024, the World Artificial Intelligence Conference released the Shanghai Declaration on Global AI Governance, calling for global efforts to promote the healthy and sustainable development of AI. During the 19th United Nations General Assembly, China released the AI Capacity-Building Action Plan for Good and for All and proposed the establishment of the "Friends Group for International Cooperation on AI Capacity Building", outlining Five Visions and Ten Actions. The National Cybersecurity Standardization Technical Committee of China released version 1.0 of the AI Safety Governance Framework.

    With emphasis placed on innovative development, the United States has issued a number of policy documents such as the Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, for the purpose of increasing support for the AI innovation ecosystem in terms of funding and talent, guiding the establishment of best practices and standards for AI risk management, and implementing export controls on advanced chips to maintain the US's leading position in AI.

    Focusing on AI regulatory legislation, the European Union has successively introduced strategies, declarations, plans, and guidelines related to AI. In 2024, it launched the EU Artificial Intelligence Act, the first ever act of this kind, to lead global AI governance.

    The United Kingdom views AI as an opportunity to enhance its global influence by issuing the National AI Strategy and adopting a "supportive of innovation" regulatory approach. In 2023, it held the inaugural Global AI Security Summit and secured the signing of the Bletchley Declaration, the first international declaration on AI security.

    Developing countries are also taking proactive actions. Kenya regards artificial intelligence and blockchain as "critical technologies supporting the economy and business", and launched the Information Technology - Artificial Intelligence - Code of Practice for AI Applications in 2024. Egypt established the National Artificial Intelligence Council in 2019 and subsequently released documents such as the National Artificial Intelligence Strategy. Kazakhstan released the Concept for Artificial Intelligence Development for 2024-2029 (Draft) in 2024, accelerating the construction of its domestic AI ecosystem.

    4. Operations of Technology Companies

    Large technology companies have competitive strengths in technology, industry, and market share. They are the main players in formulating and implementing global AI rules.

    At the organizational level, technology companies have not only established internal ethics committees but also launched corporate AI principles covering aspects such as benefiting the society, safety, privacy protection, fairness, transparency, explainability, controllability, and responsibility.

    In terms of industrial practice, technology companies have provided solutions for enhancing the security and robustness of AI systems, protecting data privacy, improving transparency, enhancing explainability, ensuring fairness and inclusiveness, aligning values, and conducting content safety audits.

    With regard to international cooperation, twenty technology companies announced at the Munich Security Conference that they will jointly combat deepfake information and strive to detect and resist deceptive AI-generated content. Additionally, 16 companies signed the Frontier AI Safety Commitments, pledging such measures as internal and external testing, information sharing, cybersecurity investments, third-party vulnerability reporting mechanisms.

    In general, consensus has been built on the general principles of global AI governance. However, the mechanisms remain highly fragmented, with significant differences in refining concepts and implementing solutions. Developed countries are home to many multinational companies in the AI field, and link AI with freedom, democracy, human rights, and other issues. The AI governance platforms led by developed countries have a longer history and greater influence. In contrast, most developing countries lack leading AI technology companies and focus more on capacity-building, AI divide, and other development issues. They have weaker voices, less representation, and insufficient discourse power in the global governance process. Furthermore, the attempts of some countries to draw ideological lines, build exclusive blocs, and create development barriers have hindered cooperation in global AI governance.

II. Ten Key Issues in Global AI Governance

    Continuous breakthroughs in algorithm models, computing power and data have further highlighted the strategic position of AI technology which has exerted significant and far-reaching impacts on economic development, social progress, and geopolitics. Currently, ten key issues have emerged in global AI governance.

    1. Principle of National Sovereignty and AI Development

     The development of AI technology is closely related to national sovereignty. On the one hand, the strategic development of AI technology can be used to safeguard national sovereignty, security and development interests. On the other hand, national sovereignty faces challenges stemming from security risks related to technological autonomy, algorithm models, data, content, computing power, and supply chains. To address these issues, the international community has conducted discussions on the applicability of the principle of national sovereignty in the field of AI, changes in the definition and scope of national sovereignty in the AI era, and ways to safeguard national sovereignty, security, and development interests.

    2. Social Change and Sustainable Development

    While AI technology empowers socio-economic development, it also poses challenges to existing social orders. Developmental issues include leveraging AI to facilitate the transformation and upgrading of traditional industries, improve production efficiency, empower smart healthcare, smart education, and smart cities, create new job opportunities, and promote green and low-carbon economic and social development. Constraints include the environmental burden caused by the vast computational resources and energy required for AI training and operation, security risks posed by AI in fields such as chemistry, biology, and nuclear energy, and structural unemployment resulting from AI development.

    3. Technological Innovation and Industrial Development

    Accelerating AI development is a strategic issue concerning whether one can seize the opportunities of the new round of technological revolution and industrial transformation. Related discussions include: AI technological research and development, namely focusing on advancing frontier basic theories, key common technologies, and enhancing intelligence levels; AI computing power which entails accelerating the deployment of computing infrastructure, optimizing the layout of computing resources, and efficiently allocating and utilizing computing power to promote technological and industrial development; AI innovative applications which involves transforming innovative achievements into new productivity in various fields; AI industrial policies that address how to leverage resources such as systems, mechanisms, funding, and talent to develop AI; AI standard setting which focuses on establishing industry standards for developing, deploying, and using AI; and AI supply chain and industrial chain security which involves avoiding the use of advanced chips, critical software, and other components necessary for AI development as tools for geopolitical games among a few countries, and ensuring that while the AI supply chain and industrial chain are maximally embedded in the global division of labor, they are also resilient and capable of mitigating risks.

    4. Human-Machine Emotions and Bioethics

    AI technology may raise fundamental ethical issues that threaten human survival and development. The global community primarily focuses on the following issues: exploring emotional attachment in human-AI interaction, addressing emotional illusions in human-AI interactions, and preventing interpersonal alienation and system addiction; ethical decision-making and moral judgment in AI, including establishing AI ethics rules to prevent AI from losing control or harming human life, rights, and interests; and the structural impacts of AI, which may disrupt views on employment, fertility, and education.

    5. Content Security Risks

    False and misleading information generated by the misuse of AI technology has   exacerbated content security risks. Relevant governance issues include:large models potentially generating incorrect, inaccurate, or unrealistic "hallucinated" content due to their tendency to imitate based on training data rather than the ability to understand; malicious users synthesizing realistic images or audio-visual content using deepfakes and other technologies to manipulate public opinion and spread false information; and intelligent algorithm recommendations catering to users' preferences, potentially creating or exacerbating "information cocoons" that narrow individual cognition and deepen stereotypes.

    6. Model Algorithm Security Risks

    Model algorithms are core components of AI systems, directly influencing system operation. They primarily involve: the interpretability of model algorithms or the need for predictable, attributable, correctable, and accountable output results; the reliability of model algorithms, which necessitates avoiding biased or discriminatory outputs and "intelligent hallucinations"; the robustness of model algorithms to ensure that they are resistant to performance degradation and decision-making errors caused by complex, multilateral operating environments, malicious interference, or inducements; the fairness of model algorithms to ensure that the decision-making process does not favor certain groups or individuals to prevent discrimination and unfairness; and the attack resistance of model algorithms to protect against risks such as theft, tampering, backdoor embedding, and attacks on core information concerning parameters, structures, and functions.

    7. Data Security and Privacy Protection

    Data security is crucial for ensuring safety, reliability, controllability, and fairness in AI development. Global concerns in data security primarily include: avoiding unauthorized collection and improper use of personal data and information during data collection and use; ensuring data quality and preventing the use of false, biased, or intellectual property-infringing information in data training, as well as guarding against data tampering, contamination, and irregular data labeling; and preventing data leaks and unauthorized access, malicious attacks, and induced interactions during data processing.

    8. Product Liability and Risks

    A major challenge in AI governance is that AI products (such as embodied intelligence) involve multiple complex technologies and diverse legal entities, making it difficult to ascertain liability when issues arise. The challenges include: the liability of AI product designers or operators for errors, negligence, or intentional misconduct in models or algorithms; the liability of AI product manufacturers for product defects or failure to meet expected safety standards; and the liability of AI product users for improper operation.

    9. Intellectual Property Protection

    AI technology and related product applications pose new challenges to intellectual property protection which require ongoing adjustments by policymakers and legal drafters. In terms of patents, AI algorithms, models, and certain components of systems may be eligible for patent protection, but pure algorithms or mathematical methods are not patentable in some jurisdictions. Regarding copyrights, no consensus has been reached among countries on whether AI-generated content can be protected by copyright, and there is controversy over whether rights belong to the developer, user, or other relevant parties of AI products. In terms of infringement, AI systems using copyrighted data for training pose infringement risks, and AI-generated brand names or logos may also involve trademark protection issues.

    10. AI Divide and International Collaboration

    The issue of the AI divide among countries represents a new manifestation of the global digital divide in the context of AI technology development. There is an urgent need for enhanced international collaboration for resolving this issue. The AI divide is intensifying across policy, technology, industry, application, and governance domains. Although all countries emphasize international collaboration in AI, including the development of international standards, enhanced policy interoperability, and international research and development cooperation, some countries are imposing technological blockades in the AI field, seeking leadership by blocking the AI development of other countries. If this trend is not restrained, it will seriously disrupt the global AI development and governance process.

III. Building a Sound Global

AI Governance System

    To better seize the opportunities presented by AI development, address the global risks and challenges it entails, and enhance the effectiveness of global AI governance, further efforts should be made to improve the construction of the global AI governance system. It primarily involves clarifying objectives and purposes, fostering consensus on principles, and expanding and deepening action pathways.

    1. Objectives and Purposes

    The objectives and purposes include: strengthen information exchange and technical cooperation, jointly manage risks, form a widely agreed-upon AI governance framework and standard specifications, establish an open, fair, and effective governance mechanism, continuously enhance the security, reliability, controllability, and fairness of AI technologies, and promote AI technologies for the benefit of humanity in the efforts of building a community with a shared future for mankind.

    2. Principles and Consensus

    2.1 Respect for National Sovereignty

    Countries have the right to independently choose their technological development models and governance plans based on national conditions. When providing AI products and services to other countries, it is important to respect the sovereignty and laws of the recipient countries and oppose the use of AI technologies and applications to interfere in their internal affairs.

    2.2 Balancing Development and Security

    Global AI governance should adhere to the principle of balancing development and security. On the one hand, it is important to respect the laws of technological development, encourage and promote AI technological innovation, and unleash its application potential in various fields. On the other hand, there is need to uphold a risk-based governance philosophy and apply safety awareness and regulatory measures to all stages of the AI life cycle, including research, design, development, deployment, and use.

    2.3 Adhering to Equality, Mutual Benefit, and Inclusiveness

    The development of AI should adhere to the principles of equality, mutual benefit, and inclusiveness. First, it is important to promote equal rights, opportunities, and rules for all countries in the development and utilization of AI technologies. Second, it is important facilitate the sharing of AI technologies and knowledge, reduce technical barriers, and narrow the AI divide. Third, it is important to promote openness and competition in the global AI market, prevent monopolistic behaviors, and avoid politicizing technology issues, so that all parties in the world can share the benefits brought by AI.

    3. Action Taking Pathways

    3.1 People-Centered Approach

    Global AI governance should always uphold the people-centered philosophy that respects human rights and interests and promotes sustainable human development for improving human well-being. It should actively support the use of AI to facilitate sustainable development and address global challenges such as climate change and biodiversity conservation.

    3.2. Commitment to Intelligence for Good

    It is important to abide by applicable international law, conform to the universal values of peace, development, equity, justice, democracy, and freedom, and jointly prevent and combat the misuse and abuse of AI technologies by terrorist, extremist forces, and transnational organized criminal groups.

    3.3 Empowering Various Industries

    It is important to collaborate on AI model research and development to promote the optimization and upgrading of key AI technologies and system platforms, improve the layout of globally interoperable AI and digital infrastructure, and provide computing power and algorithm resources for countries, especially developing ones, to engage in technology application and scenario innovation. It is also important to accelerate the deep integration of the AI innovation chain and industrial chain to form a diverse and healthy AI development ecosystem for good, promote all-round, full-chain, and multi-scenario empowerment of AI in the real economy, and support sustainable development, climate change adaptation, and biodiversity conservation through AI.

    3.4 Preventing and Managing Security Risks

    It is important to adopt a proactive and agile response strategy, closely monitor changes in AI security risks, and make quick, dynamic and precise adjustment to governance measures. Countries should strengthen AI data security, cooperate to promote the orderly and legal cross-border flow of data, and work for the establishment of a global mechanism and platform for data sharing. It is necessary to promote equality and diversity in AI data corpora, eliminate racism, discrimination, and other forms of algorithmic bias, and protect cultural diversity, support the establishment of a globally interoperable AI security risk assessment framework, standards, and governance system under the United Nations framework that takes into account the interests of developing countries. Countries should jointly assess AI research and application risks, improve technologies and policies to address AI security risks, and resolutely crack down on criminal acts involving the misuse of AI.

    3.5 International Cooperation Capacity Building

    On one hand, the global community should leverage the key coordinating role of the United Nations mechanisms, engage in North-South, South-South, and tri-partite cooperation in AI, and jointly implement the outcomes of United Nations Future Summit. Efforts should be made to enhance the representation, participation, and voice of developing countries and countries lagging behind in technology, oppose the formation of exclusive organizations that maliciously obstruct technological development in other countries, and reach international agreements on global AI governance rules based on universal participation. On the other hand, the international community should strengthen capacity building, bridge the AI divide, actively assist the global south in developing AI technologies and services, enhance public AI literacy, safeguard the digital and intelligence rights of women and children, and share AI knowledge and experiences.

    3.6 Improving Global Governance Mechanisms

    The United Nations, as a comprehensive international organization, is the main channels for global AI governance. Efforts are needed toward the establishment of an international governance body for artificial intelligence under the framework of the United Nations to coordinate major issues related to the development, security, and governance of artificial intelligence. Regional multilateral organizations and specialized organizations such as the International Organization for Standardization, the World Economic Forum, and the World Internet Conference should play crucial roles. Other organizational forms such as AI initiative mechanisms, security research networks, and corporate alliances are important participants. Interactions among political, industry, and technical organizations also need to be strengthened. Major powers should enhance political trust and cooperation willingness, establish emergency liaison and collaboration mechanisms, and properly manage competition in the AI field. Additionally, within the global AI governance system, efforts should be made to gradually promote the formation and improvement of mechanisms for the sharing of information, knowledge, risks and benefits.

    AI technology is still in a process of rapid development, and global AI governance is in its ascendancy. It bears on the future of all humanity and becomes a topical issue for all countries. Advancing global AI governance is a long-term process. To coordinate major issues related to global AI development, security, and governance, the international community should actively promote the establishment of AI governance mechanisms, support the United Nations in playing a leading role in the governance process, strengthen North-South and South-South cooperation, enhance the representation and voice of developing countries, and quickly form an AI governance framework and standard specifications based on wide consensus. Governments should actively participate in global AI governance, abide by relevant international law and norms, and strengthen international exchanges and cooperation. Major actors, such as international organizations, businesses, research institutions, social organizations, and individuals should play roles that match their respective identities and participate in the construction and implementation of the AI governance system.

    Looking ahead, facing the existing digital and AI divides between different countries and regions, governments, the scientific and technological community, the industrial sector, and other stakeholders should work together to fully leverage the potential of AI, work together to promote the healthy development of AI, jointly maintain AI security, so as to empower humanity's shared future and build a community with a shared future for mankind.

Research Report on Global AI Governance