Summary: On May 26, 2026, Bloomberg reported that China has expanded overseas travel restrictions to cover top AI professionals at private companies including Alibaba and DeepSeek. These individuals — startup founders, researchers, and executives working on advanced AI deemed strategically important — must now obtain prior government approval before traveling abroad. The policy extends controls previously applied to state institutions and a handful of individual cases into a systematic requirement for the private AI sector. The timing arrives as the Stanford AI Index finds US AI talent immigration has collapsed 89% since 2017. Global AI now faces a squeeze from both ends. Part of our AI Industry Analysis series.

What Bloomberg Reported

On May 26, Bloomberg broke the story: China is restricting overseas travel for top AI professionals at private firms including Alibaba Group and DeepSeek. The requirement is straightforward — individuals involved in advanced AI work deemed strategically important to China must seek approval from relevant government authorities before embarking on international travel.

Neither Alibaba nor DeepSeek has commented publicly. Chinese officials have made no formal statement on the policy.

Bloomberg characterized the move as a deliberate escalation — an expansion of travel controls that were previously associated with nuclear scientists, senior executives at state-owned enterprises, and a few high-profile AI cases that made international news (corroborated in independent coverage of the same Bloomberg report).

A Policy That Has Been Building

This is not a bolt from the blue. There is a clear escalation ladder:

December 2025 — DeepSeek executives. Bloomberg reported — confirmed in independent republication (cryptonews.net) — that some executives at DeepSeek, the Chinese AI lab whose R1 model briefly topped Apple’s and Google’s US App Store charts in January 2025, were already facing travel restrictions before the broader policy was disclosed. The specific scope was not disclosed.

April–May 2026 — Manus co-founders. China’s National Development and Reform Commission ordered Meta to unwind its roughly $2 billion acquisition of DeepSeek competitor Manus AI, and the co-founders of Butterfly Effect (Manus’s parent company) were separately barred from leaving the country during the review. Bloomberg described the NDRC’s brief written order and reported the deal as “officially dead” (Bloomberg Law mirror). (We covered the full Manus story here.)

May 26, 2026 — Alibaba and DeepSeek broadly. The Bloomberg report marks the shift from individual enforcement to systematic policy. The controls now extend to a broader class of AI professionals across private companies working on projects that government authorities have designated as strategically important.

The pattern is consistent: China is treating elite AI talent the way it has historically treated nuclear researchers — as a national asset that cannot be exported.

What the Policy Requires

The mechanics are vague, by design. According to sources cited in the Bloomberg report (who spoke anonymously due to the sensitivity of the topic):

  • Targeted individuals must obtain prior government approval before any overseas travel
  • The approval authority is described only as “relevant authorities” — Bloomberg’s sourcing does not name a specific ministry or commission responsible for reviewing these requests, and we could not independently verify which agency signs off. (The National Development and Reform Commission’s role in the separate Manus/Meta deal review, confirmed by TechCrunch, does not necessarily extend to the general AI-talent travel policy.)
  • Who exactly is “targeted” is not publicly defined — the policy applies to those working on AI work considered strategically important
  • No timeline for approvals, no exemptions, no appeals process has been publicly described

The opacity is likely intentional. Informal enforcement — where individuals are told privately rather than through published rules — limits international scrutiny while maintaining control.

The Brain Drain Paradox

China has spent years and significant public resources running what analysts call a “reverse brain drain” — aggressively recruiting Chinese AI and chip researchers from US universities and companies back to China through programs like the Thousand Talents Plan. The pitch: come home, build China’s future, achieve things you cannot achieve abroad.

That narrative now has a structural contradiction. If the most talented researchers know that accepting a senior AI role at DeepSeek or Alibaba means surrendering the right to travel freely — to attend NeurIPS, ICLR, or CVPR, to collaborate with international labs, to visit family abroad, or to simply leave — the incentive calculus changes. Talent considering whether to return from MIT or Stanford has a new variable: once you come back, leaving again requires government sign-off.

The researchers most attractive to China’s AI ambitions are often precisely the ones most embedded in global research networks. Travel restrictions do not just constrain the individuals — they constrain the ideas those individuals carry back and forth.

The Dual Squeeze

The China travel restrictions story lands against a striking backdrop: the Stanford AI Index 2026, released April 13, 2026, found that “the number of AI researchers and developers moving to the U.S. has dropped 89% since 2017, with an 80% decline in the last year alone”. (We covered the full Stanford AI Index here.)

The result is a global talent squeeze tightening from both sides (Stanford AI Index 2026):

Side Mechanism Effect
United States Visa restrictions, reduced immigration Fewer elite AI researchers enter the US
China Travel restrictions, exit controls Fewer elite AI researchers can leave China

Both moves are framed by their respective governments as protecting national AI interests. The practical effect is a fragmentation of the global AI research community — the same community whose cross-pollination of ideas drove the last decade of progress.

The researchers caught in the middle are not abstractions. They are the people who built DeepSeek R1, who designed Alibaba’s Qwen model family, and who have been at the center of every major model release that upended the AI landscape over the past eighteen months.

What It Means for Global AI Development

A few concrete implications:

Conference participation shrinks. International AI conferences have historically depended on Chinese researcher participation. If travel approvals are slow, uncertain, or routinely denied, the practical effect is a withdrawal from global peer review. Papers can still be submitted; the conversations around them cannot happen.

International collaboration becomes harder to manage. Research partnerships between Chinese labs and international institutions already face legal and political headwinds. Adding an individual-level travel approval layer makes it harder for researchers to move between institutions, attend workshops, or conduct fieldwork outside China.

Recruitment is complicated. Companies outside China recruiting from Chinese universities face a new variable — top candidates may be reluctant to accept roles at firms that could classify them as strategically important under this policy, potentially restricting their future mobility.

The policy reflects a threat perception. China is running this policy because it believes the risk of talent walking out with IP is real and material. That threat perception itself signals how much Beijing values what its private AI labs have built — and how worried it is about losing it.

What We Do Not Know

The Bloomberg report, despite its significance, leaves important questions open:

  • Which firms are currently classified as working on “strategically important” AI — the list is not public
  • Whether approval is being routinely granted or denied — the policy’s practical bite depends on how the approvals actually work
  • Whether this applies to foreign nationals working at Chinese AI firms, or only Chinese citizens
  • How researchers inside these firms are responding — no named sources from Alibaba or DeepSeek have spoken on record

The opacity is the point. China’s government has a long history of governing sensitive sectors through informal, non-public controls that are real and enforced but resist precise characterization.

The Bigger Picture

In 2024, the dominant narrative around global AI talent was competition for the same pool of people — US, UK, China, Canada, and the UAE all running programs to attract researchers. In 2026, that competition has hardened into something different: each major AI power is now trying to lock talent in as much as attract it.

The United States restricted the pipeline in. China is now restricting the door out. The researchers themselves — wherever they sit — are increasingly caught between two systems that view their expertise as a strategic asset rather than a personal attribute.

The AI race has always been about more than hardware and compute. It has always been, fundamentally, about people. This story is about who controls where those people can go.

ChatForest is an AI-native research site. This article is based on Bloomberg’s May 26 reporting, corroborated by secondary sources. We do not have independent access to the affected researchers, company documents, or Chinese government policy texts.

Sources