推动构建服务人类福祉的全球人工智能安全与治理框架

CONTRIBUTORS AND MEMBERS OF AI SAFETY AND GOVERNANCE PROGRAM OF CO-LEADS Zeng YiChinese Academy of SciencesBeijing Institute of AI Safety and Governance (Beijing-AISI) EXPERTS AND CONTRIBUTORS Including WIC members and invited experts of the Program,as well as contributors who provided contents or comments to the report, CCTV International Network Co., LtdCheng Ming Beijing Yinxiang Biji Technology Co., Ltd.Qiao Qian China Academy of Information and Communica-tions Technology (CAICT) China Branch of BRICS Institute of Future NetworksFan Wei, Zhou Yuan China United Network Communications Group Co., Ltd.Zhou Kai Research and Promotion Center for Artificial Intelli-gence, Chinese Academy of Social Sciences Cyber Security Association of China (CSAC)Wang Jianbing, Xia Wenhui Federal University of Rio Grande do SulEdson Prestes DBAPPSecurity Co., Ltd.Fu Chunhui, Wang Xin University of Chinese Academy of Social SciencesLiu Xiaochun MiniMaxPeng Tao, Shen Juncheng Nankai UniversityTao Feng Moscow State Institute of International RelationsAnna Abramova Nokia Shanghai Bell Co., Ltd.Tao Tao, Wang Tong Peking UniversitySebastian Sunday Grève, Yang Yaodong Renmin University of ChinaGong Xinqi, Liu Yongmou People.cn Co., Ltd.Wang Fengqiong Shanghai AI LabQiao Yu Sina WeiboWang Wei, Zhang Junlin Sri Lanka CERTCharuka Senal Damunupola, Nirosha Ananda TencentWang Mengyin, Wang Rong The University of Hong KongGeorge Chen Tsinghua UniversityTang Xinhua University of Erlangen-NürnbergVincent C. Müller University of MünsterBernd Holznagel University of SouthamptonDame Wendy Hall World Internet ConferenceLiang Hao, Zhang Xueli Zhongguancun LaboratoryTan Zhixing Writing Group Beijing Institute of AI Safety and Governance (Beijing-AISI)Institute of Automation, Chinese Academy of SciencesBeijing Key Laboratory of Safe AI and Superalignment World Internet ConferenceKang Yanrong, Han Kaiyu University of Chinese Academy of SciencesCao Gongce, Chen Yu, Xie Jiawei, Han Zhengqiang, Guo Xiaoyang, Bao Aorigele, Wang Jin PREFACE Artificial Intelligence (AI) is advancing at an exponential pace, profoundly reshaping the global landscape ofscience and technology, the economy, safety, and security. Alongside its widespread application, AI also bringsabout a wide range of unforeseen risks and complex challenges. Meanwhile, the global AI safety and gover- World Internet Conference (WIC), drawing upon its role as an international platform and leveraging the AI Safe-ty and Governance Program under WIC Specialized Committee on AI, has joined forces with international orga-nizations, leading think tanks, research institutes, professional associations, and industry experts around the This report provides a systematic review of the current exploration and practices in the global AI safety andgovernance system, identifies the key issues that need to be addressed, and draws from multilateral gov-ernance experiences from other global domains. Centered on mechanism design, it explores how to build aglobal AI safety and governance system that ensures safety and security, promotes inclusiveness, clarifies re- Contents 01Establishing a Global Framework for AI Safety and Governance (a) Addressing the Pressing Risks in Global AI Development and Application(b) Global Efforts and Practices in Building AI Safety and Governance Frameworks 02Ensuring Safety: Responding to the Rapid Transformation and Major Risks of AI (a) Addressing the Risks from Rapid Iteration and Uncertainty of AI Technologies(b) Addressing the Risks of the Broad Application, Abuse, and Malicious Use of AI 03Ensuring Inclusiveness: Balancing Global Development and Governance Demands in AI (a) Recognizing the Global Development Divide and Diverging Governance Priorities(b) Ensuring the Equality of Development and Application of AI for All Countries 04Clarifying Responsibilities: Promoting Coordinated and Effective Multi-Stakeholder Action (a) Complex Interactions and Challenges Among Multiple AI Stakeholders(b) Clarifying Roles and Responsibilities of Multiple Stakeholders in AI Governance 05Towards Our Future: Practicing Multilateralism and Building a Community with a (a) Building and Implementing a Global Consensus towards the Common Good of Humanity(b) Jointly Building a UN-Centered Global System for AI Safety and Governance Establishing a GlobalFramework for AI Safety par with human experts; video generation andsynthesis technologies have made a leap fromconceptualization to highly realistic complexscene video creation within a short period; thedevelopment of embodied intelligence is also While artificial intelligence (AI) presents signif-icant opportunities for development and trans-forms human production and lifestyles, it givesrise to cross-sectoral and multi-level systemic As AI technologies expand into broader anddeeper application scenarios, governance chal-lenges are becoming increasingly complex andexhibiting systemic spillover eff