According to Bloomberg (September 28, 2026), China has extended travel controls to the families of key AI specialists: spouses and children of executives, researchers, and founders of private AI and semiconductor companies now require prior government approval for any departure, including short and tourist trips. For the first time, this mechanism has covered the private sector: previously it was applied only to scientists at state institutes, nuclear specialists, and state company management.

image
image

What happened

Bloomberg reported on September 28, 2026, citing sources, that China's travel restriction regime has been extended to the spouses and children of executives, researchers, and founders of private AI and semiconductor companies. Direct restrictions on the employees of private companies themselves appeared earlier, in May 2026, and included Alibaba and DeepSeek specialists. Additionally, on September 15, 2026, a State Council decree came into force, allowing the ban on citizens' departure in the event of a risk of technology leakage through export control. Formally, this is not a complete ban: each trip requires individual prior approval from Beijing. Authorities have already notified several people, and, according to sources, the list of affected employees is expected to expand.

Context

Prior to these changes, travel controls in China covered narrow categories of citizens: scientists at state institutes, nuclear specialists, and state company management. Extending them to private AI and semiconductor companies means effectively equating such businesses with strategic state structures. Analysts link the tightening with fears of brain and technology drain to the US, which intensified after Meta agreed to buy the Chinese startup Manus for approximately two billion dollars, and Chinese authorities demanded the termination of this deal. In this logic, model weights and employees are considered a single class of strategic assets, and the competition for AI talent between China and the US is interpreted as a national-level threat.

Why this matters for the industry

For the industry, this is an institutional change in working conditions, not an ordinary HR news: employees of private labs and their families are now effectively within the same control perimeter as state secret bearers. Already, overseas business trips, participation of Chinese researchers in conferences, workshops, and joint projects are becoming more complicated, and live exchange formats — visits and interviews — are narrowing. Sources do not describe significant changes at the level of models, benchmarks, or APIs: no restrictions on weights, licenses, or inference have been introduced. However, over the next few months, the probability of communication shifting to institutional channels like official blogs and technical reports instead of personal presentations is growing, which may mean a more cautious open-source release policy for affected teams. Over the next couple of years, if the course is maintained, structural isolation of the Chinese research environment is possible: fewer joint publications, less independent criticism of evaluation methodologies, and a higher share of self-reported results without external verification. For founders and companies outside China, this is a signal of escalation in the fight for AI talent between China and the US, and a reminder that international deals by Chinese private labs can be politically blocked, as shown by the Manus deal.

Why this matters for users

If you use models from Chinese teams like DeepSeek, Alibaba, or Zhipu, nothing changes for you in the runtime today: APIs, weights, and clouds continue to work, no new restrictions on licenses or model output are described in the sources, and there is no need to rebuild your working stack. Practically, something else is noticeable: the number of public appearances and open benchmarks from Chinese researchers may gradually decrease, so the freshness of weight versions, license terms, and release channels should be checked more actively than before. The news is also useful to consider as background when choosing model providers: it reflects growing geopolitical tension around AI talent between China and the US, which affects the availability and openness of Chinese models in the long term.

What is still unknown / limitations

The causal link between Meta's purchase of the Manus startup and the tightening of the regime is an analyst version citing sources, not an established mechanism, so confidence in it should not be overstated. The full list of affected companies and roles, the exact number of notified employees, and the criteria for inclusion in the lists are unknown. Forecasts of a more cautious release policy, a reduction in appearances, and structural isolation of the Chinese research environment are interpretations that require verification as new data emerges.

Sources

Author

Look at AI, editorial team