China has begun forcing its top private-sector AI talent to obtain government approval before traveling overseas, a shift Bloomberg reported on May 26 that pulls firms like Alibaba Group and DeepSeek inside a control regime previously reserved for state-affiliated researchers. The move converts soft 2025-era guidance — informal advice to avoid trips to the United States — into a binding pre-approval requirement, and it lands at a moment when both sides of the US-China AI competition are increasingly worried about losing engineers as much as losing chips.
What Happened
According to a Bloomberg report citing people familiar with the matter, Chinese government agencies have started imposing overseas travel restrictions on individuals deemed to work on strategically important AI projects at private firms. The list of affected companies explicitly includes Alibaba and DeepSeek, the Hangzhou lab whose reasoning models have driven China's most visible AI breakthroughs of the past two years. Founders, principal researchers, and senior executives now reportedly need pre-trip approval from the relevant authorities before they can leave the country.
The restrictions appear to be a hardening of a softer regime that had been in place for at least six months. Bloomberg previously reported in December 2025 that some DeepSeek executives faced similar pre-travel requests, but those were treated as guidance rather than as a permit system. The expansion to a wider pool at Alibaba and across other unnamed private labs converts an ad hoc practice into a more formal channel, and it places private-sector engineers under rules that closely resemble those long applied to staff at state research institutes and military-linked universities.
Neither Alibaba nor DeepSeek has commented on the report. The Chinese government has not formally acknowledged the new procedures, which is consistent with how Beijing has historically rolled out talent-flow controls — through quiet administrative practice rather than published rules. The lack of a public legal basis is part of what makes the change difficult for companies to manage, since the criteria for being designated "strategically important" remain unwritten.
Why It Matters
For most of the AI race, the focus has been on hardware — export controls on Nvidia, foundry access at TSMC, and the ability of Chinese firms to assemble large GPU clusters. The travel curbs flip the lens onto people, which is the scarcer of the two resources. Top AI researchers can move between countries in ways that GPUs cannot, and a single hire at the right lab in the United States or Europe can effectively transfer years of accumulated training know-how out of Chinese reach. Beijing is trying to slow that channel.
The signal also matters for foreign investment in Chinese AI. Sovereign and private investors who put money into Chinese frontier labs in the past two years have done so partly on the premise that those teams can attend overseas conferences, recruit globally, and maintain credibility with the international research community. A pre-approval regime introduces a regulatory tax on all of those activities. A researcher who has to ask for a permit to attend NeurIPS will think twice about agreeing to keynote it. A founder who needs sign-off before pitching investors in New York will lean harder on local capital. The cumulative effect is a slow decoupling, not by sanction but by friction.
There is a competitive twist as well. The United States and Europe have spent the last two years tightening visas, export licenses, and outbound investment rules with the explicit goal of blocking AI talent transfer to China. Beijing's response — locking key researchers more firmly inside the country — is in some ways the mirror image. Where Washington has been trying to prevent capability from leaving the West, China is trying to prevent its own capability from leaving China. Both sides are converging on the same conclusion: in this race, the team that matters is the team you can keep.
Reaction
Western press response has tilted toward seeing the policy as confirmation of how strategic China now considers its private-sector AI labs. Coverage in the Financial Times, Bloomberg, and the Wall Street Journal has highlighted that this is the first time the formal pre-approval system has explicitly captured non-state firms at the scale of Alibaba. Industry analysts have framed it as a tacit acknowledgment that DeepSeek and a handful of peer labs are now considered as critical to national power as the state research institutes that built China's earlier AI infrastructure.
Inside the Chinese tech press, coverage has been more muted, partly because the regulation itself has not been publicly announced. Discussion has migrated to closed channels and group chats among engineers, where the practical questions are about how to interpret the new requirements: who counts as "strategically important," how long the approval process takes, what happens if a researcher needs to travel urgently, and whether students returning from overseas universities will be drawn into the same system on rehire. None of these have official answers yet.
Talent recruiters operating across the US-China corridor have noted that the announcement is likely to accelerate two trends already underway. The first is that more Chinese AI researchers who relocate abroad are choosing to formally cut ties with their Chinese employers rather than maintain dual affiliations, in order to keep their travel flexibility. The second is that hiring inside China is shifting toward candidates who have not built international networks, on the theory that they will face fewer friction points with the new regime.
What's Next
The near-term test is how Beijing handles the summer conference season. ICML in July and a series of major industry events through August will force the system to take its first round of decisions on whether to grant outbound permits to senior figures at the affected firms. If approvals come through routinely, the new regime will look like a registration process rather than a real choke point. If they slow down or get refused, the practical impact will be visible in conference attendance lists within weeks.
.jpg){kind=link}
The longer-term watch is on hiring patterns. If Chinese labs find that they cannot retain the researchers who matter most without granting them more travel freedom, the system will face internal pressure to relax. Already, several DeepSeek alumni have moved to startups in Singapore and Dubai over the past six months, partly because of the easier travel posture. Beijing is now testing whether it can hold back that drift while still letting its labs compete with US frontier teams.
Watch for a quiet response from US policymakers. Talent restrictions on the Chinese side give US institutions an opportunity to argue more forcefully for visa pathways that capture researchers wishing to leave. The CHIPS Act framework already includes some workforce provisions, and there is appetite in Congress for an explicit fast-track for AI researchers from sanctioned regimes. Beijing's pre-approval system, by making the cost of staying higher for the most ambitious researchers, may give that argument new political momentum.
Closing Thoughts
The control of AI talent is, ultimately, the control of an idea-generating workforce. Hardware can be replicated; clusters can be rebuilt; even algorithms can be reverse-engineered from published papers. What cannot be replaced quickly is a working team of senior researchers who know how to actually train a frontier-scale model. Beijing's new travel regime is an acknowledgment that the value of those teams now matches, on a per-capita basis, the value of the chip clusters they operate. That is a meaningful conceptual shift in how the Chinese state thinks about AI policy.
_sunset_2016_January.jpg){kind=link}
The risk for China is that the policy works too well in the short run and not well enough in the long run. Locking researchers in protects the leading edge today, but it accelerates the slow brain drain that happens whenever the most ambitious people in any field feel like their mobility is being constrained. The historical record on talent controls — from Soviet science cities to Cold War-era US visa restrictions — is that they buy a few years of containment and then lose value as the people most affected route around them. Whether China's AI sector is far enough ahead to absorb that drift is the real question this policy raises.
For the West, the moment is one of strategic clarity. The Chinese policy makes it transparent that AI competition is now being managed at the level of individual researchers, which changes how Washington and Brussels need to think about their own playbooks. Sanctions on chips are a blunt instrument; policies on people are a sharp one, and they are now being used by both sides. The next twelve months will tell us which set of constraints actually matters more.
한글 요약
블룸버그가 5월 26일 보도한 바에 따르면, 중국 정부가 알리바바와 DeepSeek 등 민간 AI 기업의 핵심 인력에게도 해외 출국 전 정부 승인을 의무화하기 시작했다. 기존에는 국가 연구기관과 군 관련 연구소 중심으로 적용되던 사전 승인 체계가 민간 영역까지 확대된 것이다. 적용 대상은 창업자, 핵심 연구자, 임원급으로 좁혀져 있지만, '전략적으로 중요한' AI 업무의 기준이 공개돼 있지 않아 기업들이 실제로 누구를 신고해야 하는지 판단하기 어려운 상태다.
이 조치는 단순한 출장 통제가 아니라 AI 인재 자체를 전략 자원으로 본다는 신호다. GPU 클러스터와 달리 연구자는 국가 사이를 자유롭게 이동할 수 있고, 한 명의 이직만으로도 수년치 모델 학습 노하우가 다른 진영으로 넘어갈 수 있기 때문이다. 워싱턴이 칩과 비자로 미국 외부 유출을 막아온 반면, 베이징은 자국 인재가 빠져나가는 흐름 자체를 차단하려 한다. 사실상 미·중 양측이 인재 통제라는 같은 결론으로 수렴하고 있는 셈이다.
가장 빠른 검증 무대는 여름 학회 시즌이다. 7월 ICML과 이어지는 산업 컨퍼런스에서 알리바바·DeepSeek 핵심 인력의 출국 허가가 정상적으로 나는지 여부가, 새 제도가 단순 등록제인지 실질적 차단 장치인지를 가른다. 동시에 싱가포르·두바이로 옮겨간 DeepSeek 출신 인력 사례는 통제가 강해질수록 장기적 두뇌 유출 압력도 함께 커진다는 점을 시사한다. AI 경쟁의 무게중심이 칩에서 사람으로 옮겨가는 시점이다.