Beijing has started requiring some of its most valuable private-sector AI workers to get government approval before they leave the country, according to a Bloomberg report published May 26. The people affected work at firms like Alibaba and DeepSeek, and the controls target founders, senior researchers, and executives judged strategically important to the country.
What changed isn't the existence of exit controls. It's who they now cover.
From nuclear scientists to model trainers
China has restricted the movement of people tied to sensitive work for decades, applying it to officials, defense researchers, and the management of state-owned enterprises. Extending that same logic to engineers at a privately held startup is the part that's genuinely new. Bloomberg's sources describe lists assembled not by job title but by an individual assessment of how critical a given person is to the state.
That distinction is worth sitting with. It means a mid-level researcher could be flagged while a more senior administrator is not, depending entirely on what they know and what they're building.
The reporting also notes this didn't arrive out of nowhere. Some DeepSeek executives reportedly faced similar restrictions in December 2025, months before the wider rollout. Treat that timeline as Bloomberg's secondhand sourcing rather than a confirmed government calendar, because none of this is grounded in any publicly cited statute. The enforcement is administrative and opaque, which is exactly what makes it hard to verify and easy to expand.
The Manus shadow
Hovering over all of this is the Manus case. Authorities reportedly barred two co-founders of the Chinese AI startup from leaving while regulators looked into a transaction, and last month Chinese regulators blocked Meta's planned $2 billion acquisition of the company, which had relocated to Singapore in 2025. People familiar with the matter told Bloomberg the latest travel rules aren't necessarily tied to Manus directly.
Maybe not directly. But it's hard to read "founders investigated, then barred from leaving, then acquisition blocked" and conclude the two things live in separate universes. Preventing technology from walking out the door is, by Beijing's own framing, the whole point.
So what does Beijing actually want?
The stated goals are protecting sensitive technology from leaking abroad and accelerating China's AI development against the US. The subtext is that the government now treats frontier researchers as national assets rather than free agents who can take a job in Singapore or present a paper in Vancouver whenever they like.
This isn't entirely without precedent in softer forms. Some private-sector AI staff had already been asked to report overseas travel plans, even when prior approval wasn't required. And last year the Wall Street Journal reported that authorities had advised leading AI founders and researchers to avoid visiting the US, guidance that stopped short of a formal ban. The new measures harden that nudge into a checkpoint.
The risk nobody's pricing in
Neither Alibaba nor DeepSeek commented, and the Ministry of Industry and Information Technology didn't respond to a request for comment either. There was no immediate market reaction, which tells you investors haven't figured out how to model this yet.
Here's the quieter cost. If top researchers start seeing these restrictions as a cap on their careers, the firms most central to China's AI ambitions could face a slow bleed of talent, or at least a chill on it. The signals to watch aren't dramatic. They're researchers quietly leaving, a dip in published output, fewer Chinese names at the big international conferences. None of that shows up on a balance sheet.
For now the policy exists in the space where Beijing has neither confirmed nor denied it. Watch the conference rosters this summer for the first real evidence of who can still get on a plane.
