China is making it harder for its top AI minds to leave. According to TechCrunch AI, researchers, startup founders, and executives at private firms now face travel restrictions, and some of the country’s most prominent AI figures reportedly need government approval before heading abroad. This is significant because it signals a shift in how Beijing treats AI talent: not as a free market resource, but as a strategic asset to be guarded like rare earths or military tech.
The move didn’t come out of nowhere. TechCrunch AI notes that back in March 2025, the Wall Street Journal reported Chinese authorities were already advising top founders and researchers to avoid trips to the U.S. The clampdown appears to have tightened around the Manus-Meta deal. Beijing barred Manus’ two co-founders from leaving the country while regulators investigate whether Meta’s $2 billion acquisition violates foreign investment rules. The founders are reportedly weighing how to unwind the deal, including raising roughly $1 billion to buy the company back.
Why this is happening now
The timing tracks with how close the AI race has gotten. Stanford’s latest index shows the performance gap between the top U.S. and Chinese models shrank to just 2.7% as of March 2026, down from about 31% in 2023. The U.S. still leads on model quality and high-impact patents. But China is closing fast, and in some areas pulling ahead, on publications, citations, and raw patent volume.
When you’re that close to the front, you stop letting your best people wander off to a competitor’s lab. That’s the logic here. Talent has become the bottleneck for training and fine-tuning frontier models, and Beijing is treating its researchers the way it already treats sensitive materials.
Part of a bigger pattern
Travel limits are just one piece. TechCrunch AI, citing Bloomberg, reports China also plans to police U.S. capital flowing into its top AI firms. Companies like Moonshot AI, StepFun, and ByteDance would need government sign-off before accepting American money.
Stack that on top of the economic countermeasures from 2025:
- Two rounds of export controls on 14 rare earth materials critical to high-tech military manufacturing.
- A ban on state-funded data centers deploying foreign AI chips.
What stands out is the direction of travel. The U.S. has spent years restricting chip exports to China. Now Beijing is building its own walls, around people, around capital, around materials. The decoupling is no longer one-sided.
What it means looking ahead
Expect two AI ecosystems that talk to each other less and less over the next one to three years. If Beijing keeps researchers home and screens foreign investment, the cross-pollination that defined the last decade of AI, shared papers, shared talent, shared funding, starts to dry up. That changes how fast ideas spread and where breakthroughs land.
For businesses and practitioners, a few practical takeaways:
- If you hire AI talent globally, assume the pool of mobile Chinese researchers shrinks. Recruiting pipelines that ran through Chinese labs or U.S.-based Chinese nationals get more complicated.
- If you invest in or partner with Chinese AI firms, budget for regulatory friction. Deals like Manus-Meta show approvals can be reversed after the fact, not just blocked upfront.
- If you build on frontier models, plan for two parallel stacks. A Chinese model ecosystem and a Western one, with different chips, different funding, and different rules, is becoming the base case rather than the exception.
The larger story is that AI has crossed fully into national security territory. Countries no longer treat models as products. They treat them as power. And when something becomes that valuable, governments stop letting it move freely, whether it’s silicon, capital, or the people who build the systems.
China catching up to within 2.7% explains the urgency. The next question is whether locking talent in actually keeps an edge, or whether isolation slows the very researchers Beijing is trying to protect. You can read the full reporting at the original source.