China’s internet regulator removed more than 14,000 AI products from the country’s networks in the opening phase of a sweeping cleanup campaign called “Qinglang”. The move signals a major tightening of domestic AI control.

Here is what the crackdown covered, how tech giants responded, and what the next phase targets across the sector.

What China’s Qinglang AI Cleanup Actually Did

Qinglang, meaning “Clear and Bright,” is an annual internet governance campaign run by the Cyberspace Administration of China (CAC) to remove harmful or illegal online content. The 2026 edition focused heavily on AI across the entire ecosystem for the first time.

The scale of the first phase was significant. The CAC removed over 14,000 non-compliant AI products, including websites, apps, and AI agents. Furthermore, it scrubbed more than 6 million pieces of illegal or harmful information across Chinese networks.

Follow us on X to get the latest news as it happens.

The regulator also went after accounts and datasets. It suspended over 26,000 accounts and took down more than 1,300 AI-related product listings. Moreover, it removed nine open-source datasets that it deemed illegal under current Chinese rules.

The campaign began in April 2026 and targeted four main problems. These included skipping mandatory model registration, weak safety filtering, AI data poisoning, and content not properly labeled as AI-generated across platforms and services.

New obligations now apply across the board. AI services must register, implement safety filters, clearly label AI-generated content, and properly manage training data. Failing to comply can now trigger real takedowns and penalties for offending companies.

How Tech Giants Responded to the Crackdown

Major Chinese technology firms moved quickly to comply. Huawei added special reviews in its app store, while Alibaba improved its content identification systems. Zhipu built a new review model, and DeepSeek added checks to stop data manipulation.

Meanwhile, ByteDance’s Doubao and the Qwen team took a different route, disabling their custom agent features rather than meeting the new anti-addiction and instant-exit requirements.

Local internet offices also adapted their approach. Beijing paired platform self-checks with routine monitoring, while Shanghai tailored rules by platform type.

Zhejiang focused on model auditing, and Guangdong built a multi-agency mechanism across the full AI chain.

The second phase raises the stakes further. It will target AI used to spread disinformation, produce violent material, impersonate people, harm minors, and run paid astroturfing campaigns. The regulator promised heavier penalties for offending accounts and institutions.

A separate rule also takes effect on July 15. The Interim Measures for AI Anthropomorphic Interactive Services target AI companions built for emotional relationships. It bars virtual-companion services for minors and requires guardian consent for users under 14.

The crackdown lands amid intense US-China AI competition. Chinese firms now match new US systems within months of release. Notably, security firm Semgrep said a free Zhipu model recently outperformed Anthropic’s Claude Opus 4.8 at finding software vulnerabilities.

Источник: BeInCrypto