China broadens AI travel restrictions from DeepSeek to additional private companies.

      Leading researchers at private AI companies in China are being required to hand over their passports, marking a subtle increase in the restrictions that started with DeepSeek earlier this year.

      China is broadening its unofficial travel limitations on senior AI researchers, extending measures that originated at DeepSeek to a wider array of private firms operating at the forefront, as reported by Bloomberg on Tuesday, citing sources familiar with the situation. This trend has become increasingly apparent. Top engineers and researchers are being asked to give their passports to their employers, with the official rationale being that their roles could expose them to information that might be deemed a state or commercial secret. This policy is sometimes framed as a corporate rule and other times as government guidance, but in practice, the line between the two is often unclear. Bloomberg's report situates this development within the Communist Party's emerging view of cutting-edge AI as a valuable national asset.

      The first publicly acknowledged instance occurred in March when DeepSeek employees started surrendering their passports shortly after the lab's R1 model challenged prior assumptions about the computational power needed by Chinese frontier labs to meet Silicon Valley standards.

      Around the same time, The Wall Street Journal reported that Chinese authorities had cautioned leading AI entrepreneurs against travel to the US, expressing concerns that researchers might leak sensitive technical information internationally, that American firms could acquire important intellectual property, or that executives might face detention as a form of diplomatic leverage.

      What is new in this week’s Bloomberg report is the extended reach of these controls. They now affect a wider range of individuals in the private AI sector, going beyond DeepSeek and the immediate aftermath of the R1 model release.

      These travel restrictions coincide with increasing financial regulations. In late April, China's National Development and Reform Commission and several other agencies directed major AI companies, including Moonshot AI, StepFun, and ByteDance, to turn down US-based investments in forthcoming funding rounds unless they receive prior approval.

      Multiple Chinese AI startups, including Moonshot, are contemplating shifting their corporate registrations from foreign jurisdictions back to mainland China, following Beijing's blocking of Meta’s $2 billion acquisition of Manus and indications that offshore entities would encounter more stringent domestic IPO approvals. The intent is clear: to consolidate capital, talent, and corporate presence within Chinese borders.

      The urgency behind Beijing's actions is reflected in public performance metrics. According to Stanford's 2026 AI Index, the gap between the top US and Chinese models has narrowed to 2.7%, down from 17.5 to 31.6 percentage points in mid-2023. China now accounts for 69.7% of global AI patent filings, produces 23.2% of the world’s AI publications, and deploys industrial robots at nine times the rate of the US.

      Since 2017, the migration of AI talent to the US has decreased by 89%. The combination of a shrinking capability gap and a continuous inward concentration of talent provides context for Beijing's passport policy.

      However, this policy has its drawbacks. The requirement to surrender passports functions informally as an exit ban without legal oversight, complicating international collaboration, which has historically been a strength of Chinese AI research. Whether private-sector researchers will accept this trade-off remains to be seen.

      The enforcement of travel restrictions proves easier in theory than in practice, especially as the affected group expands from a limited number of DeepSeek staff to thousands of researchers across the broader Chinese AI landscape.

      As of Tuesday evening Beijing time, neither DeepSeek nor Moonshot AI had commented publicly on the latest developments.