China Considers Restricting Overseas Access to Advanced AI Models

Chinese authorities are considering restricting overseas access to advanced domestic AI models, holding meetings with firms like Alibaba, ByteDance, and Z.ai. The move, which mirrors U.S. national security controls on its own AI, could make AI tech theft punishable under security law and follows concerns about U.S. models like Anthropic's Mythos being used against Chinese interests.

Chinese AI export restrictions

• ▪The discussions, led by China's Ministry of Commerce, covered potential limits on both closed-source and more open-weight AI models
• ▪Chinese authorities have held meetings with top tech firms about potentially restricting overseas access to China's most advanced AI models
• ▪Any new restrictions on Chinese AI models may only apply to future models, and it is not yet clear when or if they would be implemented
• ▪Companies present at the meetings with Chinese authorities included tech giants Alibaba and ByteDance, as well as the startup Z.ai

National security penalties

• ▪Officials discussed making any leak or theft of proprietary AI technology a punishable offense under China's national security law
• ▪Chinese officials also raised the possibility of new measures to restrict who can fund domestic AI startups

US AI access controls

• ▪In June, the Trump administration ordered that foreign nationals not have access to Anthropic's advanced Fable and Mythos models
• ▪The U.S. has also restricted access to its top AI models, citing national security concerns about their potential misuse
• ▪Anthropic's Mythos model, designed for cybersecurity professionals, is still only available to some "trusted" U.S. organisations

Chinese cybersecurity concerns

• ▪Chinese authorities are worried about the potential for Anthropic's Mythos model to exploit software vulnerabilities against Chinese interests
• ▪Zhou Hongyi, founder of cybersecurity firm 360, has said that China needs to develop its own version of the Mythos model

Recent Chinese AI protections

• ▪In early June, China issued new rules tightening control of overseas deals involving Chinese investors, technology, data, and national security
• ▪China has launched investigations into local AI startups like Manus that have moved abroad to determine if they broke export control laws
• ▪In April, China's state planner ordered Meta to unwind its $2 billion acquisition of the Chinese-founded AI startup Manus

Proposed regulatory framework

• ▪A May roundtable of Chinese legal experts proposed a tiered system for regulating open-source AI
• ▪The proposed tiered system includes simple filing for basic tools, security reviews for advanced tech, and barring or restricting frontier models