Introduction

In recent years, China’s artificial intelligence (AI) has rapidly developed, with its value becoming increasingly apparent. General Secretary Xi Jinping emphasized the need to comprehensively advance AI technological innovation, industrial development, and application empowerment, while improving the regulatory system for AI.

The 14th Five-Year Plan outlines the establishment of an efficient and convenient access mechanism that adapts to new business models, exploring new regulatory methods such as sandbox regulation and trigger-based regulation. Currently, China’s AI development is among the global leaders, deepening the integration of “AI +” to empower economic and social development and enhance governance capabilities.

To implement the important speech by General Secretary Xi Jinping and the decisions of the Party Central Committee, it is essential to balance development and safety, actively explore sandbox regulation, and promote the healthy and orderly development of AI in a beneficial, safe, and fair direction through effective regulatory governance.

What is Sandbox Regulation?

Sandbox regulation is a novel regulatory concept and approach that defines a specific scope within which regulatory authorities adopt inclusive and prudent measures. It involves full-process monitoring of the operations of entities within this scope, allowing for error correction and fault tolerance within a controlled environment, thus preventing the spread of issues and conflicts.

Benefits of Exploring Sandbox Regulation

Exploring sandbox regulation can create a safe and controllable “testing ground” that provides ample space for relevant enterprises to test new products, operate new business models, and enhance technological innovation in real market environments. It compensates for the shortcomings of conventional regulation in balancing innovation, benefits, and risks. Additionally, by imposing restrictive conditions and control measures, it effectively prevents the potential spread of problems.

It is important to note that sandbox regulation does not imply a lack of oversight; rather, it is conducted under the premise of maintaining safety. This requires strict access controls, ensuring that illegal and high-risk activities are not allowed into the sandbox. Monitoring must be strengthened, with regulatory authorities following up in real time, promptly correcting deviations, and decisively halting operations when necessary. Projects that mature can be promoted, while those with significant issues or uncontrollable risks should be terminated and removed from the sandbox to ensure that risks do not spill over.

Key Areas for Focus

1. Defining Regulatory Scope and Dynamic Adjustment
   AI technology evolves rapidly, involves a wide range of fields, and features new business models. Therefore, there must be clear and operable institutional arrangements for exploring the boundaries of sandbox regulation. This can be achieved by adhering to a classification and grading principle, setting more cautious regulatory standards and processes for high-risk areas such as data security and financial safety. For more mature technologies with limited spillover risks, conditions can be moderately relaxed to concentrate limited regulatory resources on key subjects.

   Regulatory subjects should be dynamically adjusted, with real-time monitoring of the operational status and risk characteristics of projects within the sandbox. If operations are stable and risks are controllable, they can transition to conventional regulation with ongoing tracking. If data exceeds normal ranges or risk levels are exceeded, timely termination is required to prevent risk spillover.

2. Innovating Regulatory Tools and Optimizing Methods
   The advancement of sandbox regulation should flexibly adopt regulatory tools and methods that align with the development patterns of AI technology and industry. Enhancing digital intelligence technology can enable panoramic real-time monitoring, adaptive control, and precise guidance for entities within the sandbox, leveraging the advantages of sandbox regulation in proactive prevention, dynamic adjustment, and transparency.

   Promoting precise and flexible regulation involves developing practical regulatory methods and differentiated strategies based on the business models, credit levels, and risk grades of innovative entities within the sandbox.

3. Improving Regulatory Mechanisms and Enhancing Effectiveness
   Establishing a scientific, reasonable, and flexible regulatory mechanism is crucial for achieving a positive interaction between the innovative development of AI technology and industry and governance compliance. A flexible fault tolerance mechanism can be established to exempt or lightly penalize entities that make mistakes or do not achieve expected outcomes during trial processes. Companies should be allowed to set their own testing periods based on technological innovation needs, reserving space for algorithm optimization, hardware upgrades, and risk adjustments.

   Strengthening service guidance in areas such as market access and financial support can prioritize promoting compliant and creditworthy AI industry projects. Providing guidance during key stages such as algorithm development, information processing, and performance verification can create a regulatory environment that encourages innovative exploration and precise fault tolerance.

   Optimizing the evaluation and feedback mechanism involves regularly conducting comprehensive assessments of projects within the sandbox and adjusting regulatory directions and measures based on feedback results, establishing a governance system for the AI industry that covers the entire process, chain, and cycle.