China moves to boost AI oversight with focus on safety, data protection

China's proposed AI law aims to boost research, improve ethics, strengthen risk checks and enhance safety oversight for AI technologies and infrastructure

China’s top legislature is tightening rules around artificial intelligence (AI) safety and ethics through a proposed amendment to its cybersecurity law, according to a report by South China Morning Post. The move reflects Beijing’s growing focus on AI governance amid rising concerns over misinformation, privacy breaches and online fraud.

 

Wang Xiang, spokesperson for the Legislative Affairs Commission of the National People’s Congress (NPC) Standing Committee, said on Thursday that the amendment draft was first reviewed last month. Lawmakers agreed that it would serve as a “guiding principle for work related to internet safety” and include new sections on AI regulation and personal data protection, the news report said.

 

 

Why China is updating its cybersecurity rules

 

According to Wang, since the cybersecurity law was first enacted in 2016, the rapid rise of new technologies has brought fresh risks and illegal online activities. The amendment aims to tackle these challenges by introducing a framework for AI security and development.

 

The proposal includes support for research in fundamental AI theories, key technologies and basic infrastructure. It also focuses on improving ethical standards, strengthening risk monitoring and enhancing oversight of AI-related safety.

 

In China’s legislative process, a draft law typically undergoes three reviews before being passed, though under certain conditions it may be approved earlier. The NPC Standing Committee will deliberate on the amendment in its session starting Friday and ending Tuesday.     

 

Beijing tightens rules on AI content

 

The new legislative push follows a series of steps by the Cyberspace Administration of China (CAC) to control AI-generated content. In March, the CAC introduced rules requiring both visible and embedded labels for AI-generated text, images, audio and video. Visible labels must be clear to users, while hidden identifiers, such as digital watermarks, must be part of the metadata.

 

In 2023, the CAC had also issued regulations mandating that providers of generative AI services protect personal data, ensure information accuracy, and respect intellectual property rights. Despite these rules, AI misuse continues to raise concerns.

 

UBS sees big AI opportunities ahead in China

 

While China focuses on regulation, analysts predict that its AI industry could soon become a major revenue driver. According to UBS, Chinese companies may start profiting from AI-powered “agents” as early as next year, with 2026 expected to mark the year of agent monetisation.

 

In a news report last month, South China Morning Post quoted UBS strategist Sundeep Gantori as saying that as models like DeepSeek’s R2 grow more advanced, monetisation opportunities will increase.     

 

China and US take different AI paths

 

UBS estimates that the US currently leads the AI agent market, generating between $15 billion and $20 billion in annual revenue. American firms are more accustomed to paying for enterprise-level software and AI tools, giving them an advantage.

 

Chinese companies have focused on consumer-facing AI uses, such as ecommerce and entertainment. However, tech giants like Tencent, Alibaba and ByteDance are now racing to build their own “agentic frameworks”, toolkits that power AI agents.

 

• Tencent launched its Youtu-Agent framework in September

• ByteDance unveiled its framework in July

• Alibaba introduced its Qwen-Agent in March

 

Although US frameworks remain dominant, Chinese alternatives are gaining traction. According to IBM rankings, ByteDance’s Coze Studio and Alibaba’s Qwen-Agent have each surpassed 10,000 stars on GitHub, reflecting their growing popularity.