On September 21, 2026, Democracy and Legal SystemWeekly published a foreword article authored by Cai Cuihong, Deputy Director of the Center for Global AI Innovative Governance and Professor at the Institute of International Studies, Fudan University. The article argues that AI governance should seek to maintain a dynamic balance between development and security by precisely defining the scope and targets of regulation, employing governance tools in a coordinated manner, and dynamically adjusting the pace of governance.
Image source: Issue No. 35, 2026 (No. 1263 overall), Democracy and Legal System Weekly
At present, a new wave of technological revolution and industrial transformation is gathering pace, ushering in an unprecedented period of innovation in artificial intelligence. AI is moving beyond the “digital world” into the “physical world” and becoming increasingly integrated into production, everyday life, and broader economic and social activities. At the same time, the risks inherent in and arising from AI technologies are becoming more complex, with growing concerns over technology ethics, data security, disinformation, autonomous decision-making by AI agents, and the security of critical information infrastructure. In particular, as large AI models become more capable and AI agents are deployed at an accelerating pace, AI is advancing beyond content generation and decision support toward autonomous execution, giving rise to increasingly complex forms of risk.
At the 2026 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance (hereinafter referred to as the “2026 World Artificial Intelligence Conference”), President Xi Jinping noted that the more rapidly AI technology advances, the more important it is to ensure that its development remains firmly oriented toward beneficial purposes and the well-being of humanity, to calibrate the scope of regulation and governance with greater precision, and to improve measures in a timely manner to prevent AI from getting out of control. This important observation captures a central challenge of AI governance. Insufficient regulation may allow risks to accumulate and spread, while excessive regulation may raise the costs of innovation and constrain room for development. The key to calibrating governance appropriately lies in making governance more targeted and adaptive—by precisely defining the objects of regulation, applying governance tools with precision, and calibrating the pace of governance—so as to maintain a dynamic balance between development and security.
First, regulation should be precisely targeted, with a risk-based and tiered approach. AI is not a single technology. Different technological forms and application scenarios vary significantly in the nature of their risks, the scope of their impact, and the potential harm they may cause. Foundation models, AI agents, and AI applications in sectors such as healthcare, finance, transportation, and education should not simply be subject to the same level of regulatory scrutiny. Regulatory requirements should be calibrated according to technological capabilities, application scenarios, scope of impact, and level of risk. Greater regulatory resources should be directed toward high-risk areas that could cause significant harm or systemic impacts, while a prudent and inclusive approach should be maintained toward lower-risk innovations that remain at an exploratory stage, rather than applying highly stringent rules across all scenarios. The Chair’s Statement of the 2026 World Artificial Intelligence Conference calls for “a risk-based approach and agile governance, while exploring classified and tiered management,” reflecting an approach that tailors governance measures to different levels of risk and enhances regulatory precision.
Second, governance tools should be precisely applied by coordinating legal, technical, standards-based, ethical, and other instruments. AI technologies evolve rapidly, involve considerable technical complexity, and span long application chains. No single governance instrument can adequately cover the entire process of research and development, deployment, and application. At the 2026 World Artificial Intelligence Conference, President Xi Jinping called for efforts to establish systems encompassing laws and regulations, technological monitoring, risk early warning, and emergency response. Laws should set out fundamental rules and safety boundaries while ensuring effective coordination with technical standards, ethical guidelines, and industry norms. For risks with a strong technical dimension, greater use should be made of technical measures such as model evaluation, safety testing, content labeling, behavior traceability, and risk early warning, thereby embedding governance requirements into the operation and application of AI systems. Different governance measures should be adopted for different types of risk: laws should establish clear boundaries, standards should translate requirements into greater detail, technical measures should strengthen risk identification capabilities, and ethical guidelines and industry self-regulation should reinforce the responsibilities of relevant actors. Governance tools should therefore be selected according to the characteristics of specific risks and combined to generate effective governance synergies.
Third, the pace of governance should be carefully calibrated, with agile governance and dynamic adjustment employed to keep pace with technological iteration. AI technologies evolve rapidly, while the development of institutional rules requires observation, deliberation, and practical experience. Rules that are fixed too early may constrain innovation, while governance interventions that come too late may miss critical windows for risk prevention. The appropriate level of governance should therefore not be treated as a fixed value over the long term, but should be dynamically adjusted in line with technological maturity, the scope of application, and the level of risk. Pilot programs, standards and guidelines, risk assessments, and other approaches can be used to identify emerging risks in a timely manner, refine existing rules, and strengthen the governance system’s capacity to respond to technological change, thereby fostering a positive interaction between institutional innovation and technological innovation.
Ultimately, precision governance should foster a positive interaction between development and security. At the 2026 World Artificial Intelligence Conference, President Xi Jinping emphasized that China has consistently attached equal importance to development and security, sought to understand the trends and underlying dynamics of AI development, and continuously improved relevant laws and regulations, policies and institutional frameworks, application norms, and ethical guidelines to ensure that AI remains safe, reliable, and controllable—so that this “swift steed” of AI can run both fast and steadily. Running “fast” requires fully unleashing the vitality of innovation, while running “steadily” requires a solid foundation for security. Together, these constitute essential conditions for the sound development of AI. As AI enters a new stage of accelerated evolution, further efforts should be made to make regulation and governance more evidence-based and refined. A risk-based approach should be adopted to precisely identify the targets of regulation; coordinated governance should be employed to ensure the precise application of governance tools; and agile governance should be used to carefully calibrate the pace of governance. In this way, sufficient space can be preserved for innovation while necessary boundaries are established for security, helping ensure that AI continues to develop in a sound manner and remains oriented toward beneficial purposes and the well-being of humanity.
Original Link: https://mp.weixin.qq.com/s/v0ak4ApKE0q7cn2vZUROHw?scene=25&sessionid=#wechat_redirect
