Introduction

The deep integration of artificial intelligence (AI) with the real economy and the cultivation of a robust intelligent industry are crucial measures for accelerating the development of new productive forces and building new national competitive advantages. The key to the “AI +” initiative lies in using cutting-edge disruptive technologies as an engine to comprehensively reshape production factors and their combinations, thereby stimulating new economic growth momentum.

AI Development in China

The 14th Five-Year Plan explicitly proposes the comprehensive implementation of the “AI +” initiative. This indicates a rapid transition of AI from a “new technology” to a “new infrastructure” across society, becoming more integrated into the production system. Driven by both policy and market forces, China’s AI industry has maintained a growth rate of over 20% for several consecutive years. By March 2025, there were 346 generative AI services registered nationwide, showcasing tremendous application potential.

Applications of AI

Deepening and expanding “AI +” is crucial for promoting beneficial, reasonable, and “good” applications of AI, ensuring that all people share in the development outcomes and better serve the modernization of China. In emergency rescue, quadruped robots can traverse rugged terrain to perform intelligent inspections and geological disaster rescues. In modern agriculture, AI-driven weather forecasting models can accurately predict hourly weather for the next 15 days, combined with smart drones and water-fertilizer management systems, significantly reducing labor costs and increasing crop yields. In terms of public welfare, intelligent rehabilitation robots can assist the elderly in daily activities and use emotion recognition algorithms to simulate emotions, providing high-quality companionship.

Challenges of AI

AI empowers various industries, bringing unprecedented development opportunities but also more complex governance challenges. For example, as model capabilities become more accessible and toolchains proliferate, the thresholds for AI “abuse” and “malicious use” have lowered, leading to compounded cross-domain risks. In cybersecurity, the landscape has shifted from “human-to-human” to “machine-to-machine” attacks, with adversaries using large models to automatically generate phishing emails, malicious code, and penetration scripts, enhancing the concealment of attacks. Furthermore, the probability of “loss of control” increases, leading to safety challenges characterized by “low probability, high damage” scenarios. Given that models heavily depend on data quality and training distributions, extensive use of machine-generated data over time may lead to model collapse and a gradual deviation from real-world representations. When many organizations use a few foundational models, single-point errors can be amplified through algorithmic replication and market interactions, creating new security risks. Although we are still distant from true “autonomous consciousness,” the risks of AI losing control remain.

Governance of AI

Governance methods are specific approaches to strengthening AI governance. Globally, governance methods vary by country due to differing governance principles. The United States advocates for relatively lenient “soft law” to provide exploratory space for AI innovation, while Europe emphasizes “hard law” to mitigate potential risks associated with AI. AI governance is essentially a comprehensive issue involving multiple stakeholders and covering various scenarios; only by categorizing governance topics can suitable mechanisms be identified for specific issues.

In 2017, China released the “New Generation Artificial Intelligence Development Plan,” and after years of exploration, has established a governance system covering key areas such as data security, content security, model security, and network security, including laws, policies, standards, and ethical guidelines. This year’s “Government Work Report” explicitly calls for improving AI governance, emphasizing the need to balance development and safety, adhere to a governance approach that combines classified and graded regulation with inclusive and prudent innovation, and further enhance governance capabilities and systems.

Recommendations for AI Governance

1. Agile Governance: Improve the legal system for technology by accelerating research and promoting flexible mechanisms like “regulatory sandboxes” to provide a dynamic and inclusive governance environment for technological iteration. In an increasingly prosperous open-source ecosystem, clarifying the legal responsibilities of participating entities is key to enhancing the legal framework for technology and promoting foundational technological innovation. By encouraging innovation while allowing for failures, an inclusive and prudent regulatory model can help AI grow into a pillar industry of the future.

2. Integration with Traditional Governance: As AI deeply penetrates various aspects of production and life, its governance must align with industry attributes. Tailored regulatory measures should be developed for different industries, establishing non-negotiable bottom lines. In critical sectors such as healthcare, finance, and transportation, AI safety risk prevention must be deeply integrated with existing laws and regulations, promoting healthy development while solidifying the bottom line.

3. AI as an International Public Good: Shape AI as a benefit for humanity by implementing the “Global AI Governance Initiative” and actively engaging in international cooperation on AI. Through the practical application of “AI +” innovation in governance, efforts should be made to establish widely recognized governance models that contribute to bridging the global AI development gap. Actively participate in international rule dialogues, uphold equal rights, opportunities, and rules, and promote the establishment of a global AI governance framework, creating a more open, fair, and effective international governance mechanism for AI.