Chinese government agencies have begun requiring top artificial intelligence researchers, startup founders, and executives at private firms — including Alibaba and DeepSeek — to obtain official approval before traveling abroad, according to Bloomberg's reporting confirmed by multiple sources May 26. The move extends a form of state control previously reserved for nuclear scientists and executives at state-owned enterprises into China's most commercially valuable private-sector technology companies, marking a significant escalation in Beijing's effort to treat AI expertise as a national security asset on par with weapons technology.
The restrictions are not rooted in formal legislation. Government agencies are applying them as administrative pressure, evaluating individual researchers based on their assessed strategic importance to China's technology ambitions — not their seniority or job title. Those affected include a mix of startup founders, senior researchers, and executives whose work is judged relevant to national AI priorities.
Neither Alibaba nor DeepSeek has commented publicly on the restrictions. China's Ministry of Industry and Information Technology has not responded to press requests for comment. Beijing has neither confirmed nor denied the policy.
DeepSeek Passport Controls Began in December 2025
The current expansion builds on restrictions first quietly applied in December 2025 to some DeepSeek executives, according to Bloomberg and TechCrunch reporting. At that time, DeepSeek's parent company High-Flyer was reported to be holding certain employees' passports rather than issuing a formal government travel ban — a softer enforcement mechanism that has since hardened into mandatory pre-approval.
The change from advisory to mandatory is material. As recently as March 2025, the Wall Street Journal reported that Chinese authorities were advising top AI founders and researchers to avoid traveling to the United States, treating it as guidance rather than a requirement — with executives instructed to report their plans before leaving and to brief authorities on their return. By May 2026, that guidance has hardened into a formal pre-approval requirement, applicable regardless of destination. DeepSeek founder Liang Wenfeng had already declined an invitation to an AI summit in Paris in February 2026.
The most documented cases involve Manus, an AI startup that relocated from Beijing to Singapore in 2025. When Meta announced a $2 billion acquisition of Manus in December 2025, Chinese regulators opened an investigation. In March 2026, the Financial Times reported that Manus CEO Xiao Hong and chief scientist Ji Yichao had been barred from leaving the country by the National Development and Reform Commission. On April 27, the NDRC formally blocked the acquisition, ordering the two parties to unwind a deal that had already substantially closed. Bloomberg sources state the Alibaba and DeepSeek restrictions are a separate, broader policy — not a direct consequence of the Manus case — but the pattern is unmistakable.
Why Beijing Escalated Now: Stanford's 2.7% Gap
The timing reflects a precise strategic calculation. According to Stanford University's 2026 AI Index, released April 14, the performance gap between the top American and Chinese AI models has collapsed to 2.7%, measured on the Arena Leaderboard benchmark — down from a range of 17.5 to 31.6 percentage points in 2023. China leads the US in AI patent output, academic publication volume, and industrial robot installations. The US retains advantages in private investment ($285.9 billion versus China's $12.4 billion in 2025) and frontier model performance, but that margin is eroding at a pace that has alarmed policymakers on both sides.
The convergence has changed Beijing's risk calculus. When China trailed the US by double digits on every capability benchmark, losing a researcher to an American firm represented a net transfer of capability from a weaker position to a stronger one. Now that Chinese models are operating at near-parity, the strategic logic has inverted: the same researcher who might once have moved to a position of relative weakness abroad is now moving from a position of strength. Beijing appears to have drawn the obvious conclusion.
The restrictions also arrive days after Huawei unveiled its LogicFolding chip architecture at the IEEE International Symposium on Circuits and Systems in Shanghai on May 25, presenting a pathway to 1.4-nanometer-equivalent transistor density by 2031 without the extreme ultraviolet lithography machines China cannot access due to US export controls. Taken together — talent controls, the Manus acquisition blocked, Huawei's chip roadmap, two rounds of rare earth export restrictions in 2025, and the NDRC's April directive barring AI firms including ByteDance and Moonshot AI from accepting US investment without government sign-off — the picture is of a unified strategy: advancing domestic capability while locking down the human capital and intellectual property that makes that capability possible.
What China's National Intelligence Law Requires of Private Companies
One dimension of this story does not depend on anonymous sourcing or policy interpretations. China's 2017 National Intelligence Law, Article 7, states that all organizations and citizens must "support, assist, and cooperate with national intelligence work." Carnegie Endowment for International Peace research notes this provision, alongside the 2021 Counter Espionage Law, appears to authorize the government to compel Chinese companies to share data with national security agencies — including from overseas subsidiaries. Legal scholars debate the precise enforcement scope: China Law Translate's Jeremy Daum has argued the provision lacks an explicit enforcement mechanism and requires intelligence work to be conducted "in accordance with law." The US Department of Homeland Security has taken the more expansive interpretation, warning that Chinese companies may be required to secretly share data on request, even when such sharing would be illegal in the jurisdiction where the data was collected.
The debate over the law's scope matters for a specific reason: the travel restrictions are a domestic administrative action, but they operate within the same legal ecosystem. The government asserting the right to approve an AI researcher's conference attendance is the same government whose statutory framework requires private firms to cooperate with intelligence work. The two instruments serve the same strategic goal — preventing technology and expertise from flowing outward in ways that could benefit adversaries.
How China's AI Talent Policy Affects Global Research and Investment
The restrictions carry practical consequences for the global AI supply chain that extend well beyond the researchers directly affected. Nvidia, TSMC, and other firms embedded in global semiconductor and AI hardware networks depend on relatively open flows of talent and collaborative research. International AI conferences — NeurIPS, ICML, ICLR — have historically served as a primary mechanism through which Chinese researchers exchange ideas with Western peers, forming the relationships that underpin global research networks. Required government approval for attendance transforms that activity from a routine professional decision into a bureaucratic and potentially career-defining one.
For investors in Chinese AI companies, analysts warn the restrictions introduce a category of risk that does not appear on any balance sheet. If top researchers at Alibaba and DeepSeek begin to perceive mandatory travel approval as a career ceiling, that could trigger a talent migration away from the most heavily controlled firms — a brain drain produced not by salary differentials but by the prospect of permanent restrictions on professional mobility. Recruitment poses a parallel problem: senior researchers trained at Western institutions are unlikely to relocate to positions carrying indefinite exit-approval requirements.
Joy Dantong Ma, associate director of MacroPolo, a think tank focused on China's economic growth, argued in an earlier analysis that a more fluid movement of scientists benefits both countries, enabling co-authorship and dialogue that produces global safety standards and accelerates basic research. "Because of the AI race mentality," she noted, "people see this as a zero-sum game." The travel restrictions suggest Beijing has decided, for now, that they are right.
Frequently Asked Questions
Why is China restricting AI researchers from traveling abroad?
Beijing views advanced AI researchers, startup founders, and executives at private firms as national security assets whose expertise cannot be allowed to flow freely to foreign competitors. The restrictions — which now require government approval before any overseas travel — aim to prevent talent poaching and the outward transfer of technological knowledge at a moment when China's AI capabilities are approaching parity with those of the United States, with Stanford's 2026 AI Index measuring the performance gap at just 2.7%.
Who is affected by China's AI travel restrictions?
Government agencies are selecting individuals based on their assessed strategic value to China's AI ambitions, not seniority or employer. Those confirmed to be affected include researchers, startup founders, and senior executives at Alibaba and DeepSeek. The two co-founders of Manus — CEO Xiao Hong and chief scientist Ji Yichao — were separately and specifically barred from leaving the country during Beijing's investigation of Meta's acquisition attempt.
What does China's National Intelligence Law require companies to do?
Article 7 of China's 2017 National Intelligence Law requires all organizations and citizens to "support, assist, and cooperate with national intelligence work." The Carnegie Endowment for International Peace and US government advisories interpret this as potentially authorizing compelled data-sharing with government agencies, including from overseas subsidiaries, though legal scholars debate the provision's practical enforcement scope.
How close is China to matching US AI capabilities?
According to Stanford University's 2026 AI Index, the performance gap between the leading American and Chinese AI models has narrowed to 2.7% on the Arena Leaderboard benchmark, down from a range of 17.5 to 31.6 percentage points in 2023. China leads the US in AI patent output and academic publication volume, while the US retains an advantage in private investment and frontier model quality.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
