China considers restricting overseas access to advanced AI models on security grounds.
Proposed AI export controls target Alibaba, ByteDance and other leading developers.
China tightens AI regulations amid intensifying technological rivalry with the United States.
China is considering restricting overseas access to its most advanced artificial intelligence models. The Ministry of Commerce led talks with major domestic technology companies over the past month to review the proposed measures, Reuters reported.
Representatives from Alibaba, ByteDance and AI startup Z.ai participated in the discussions. The potential curbs aim to restrict the distribution of both closed-source systems and open-weight models, and could also apply to unreleased future iterations, though it remains unclear if or when they will be implemented, according to two sources cited by the news agency.
The strategy mirrors an expanding effort by Beijing to classify cutting-edge AI as a strategic national asset, similar to the United States. The country is moving to secure its domestic technological framework amid intensifying global competition over artificial intelligence dominance.
Security And Legal Protections
Authorities proposed stringent penalties to protect domestic AI technology. One source said civic officials suggested making any theft or leak of proprietary AI systems a punishable offence under the country's national security legislation, Reuters reported.
The talks also explored stricter limitations on who can invest in local AI startups. This follows steps taken in June, when Beijing tightened regulations governing overseas agreements that involve Chinese investors, data, technology and national security.
Regulatory scrutiny has extended to companies relocating abroad. Authorities investigated Manus and several other Chinese AI startups to check for potential export control violations, according to three sources.
Global Market Impact
Chinese AI systems have gained significant global traction since the launch of DeepSeek's R1 model last year. Developers attract international users by providing highly capable systems at substantially lower prices than Western competitors.
Alibaba's Qwen and ByteDance's Doubao are currently among the most widely used Chinese models globally. Meanwhile, Z.ai has gained attention in Silicon Valley because its GLM-5.2 system approaches US performance levels at a fraction of the cost.
Potential export restrictions threaten to disrupt the international market. Industry observers warned that curbing overseas access could severely limit the global supply of affordable AI models.
US-China Geopolitical AI Rivalry
The proposed measures parallel actions taken by the US government. The Trump administration barred foreign nationals from accessing Anthropic's advanced Fable and Mythos models in June, which caused temporary global shutdowns because users' locations could not be instantly verified.
While controls on Fable were subsequently eased after additional safeguards were introduced, Mythos remains restricted to select trusted US organisations. According to two sources, Chinese civic officials fear the US could use Mythos to identify software vulnerabilities and target Chinese interests.
State media and prominent technology figures share these security concerns. Zhou Hongyi, founder of cybersecurity firm 360, argued that China should develop its own equivalent of Mythos
