background image blur
background image
  • Blog
    >
  • News
    >
  • China Turns AI Into a Tool for Predicting Political Risk

China Turns AI Into a Tool for Predicting Political Risk

Dominykas Zukas author photo
By Tech Writer and Security Investigator Dominykas Zukas
clock icon
Last updated: 2 June, 2026
An image of the China flag waving over a major city landscape

Key Takeaways

  • Chinese firm Geedge Networks and its government-backed research arm MESA Lab have been developing AI systems that analyze internet activity, location data, and telecommunications records to predict who may pose a political risk, according to leaked documents reviewed by Vanderbilt University researchers.
  • The system targets people who have not yet acted or spoken, building behavioral profiles designed to identify political intent before any public dissent materializes.
  • US export controls on advanced GPU chips appear to have slowed Geedge's progress, forcing the firm to rely on older, less capable hardware.
  • No evidence exists that the predictive system has been deployed, but US officials acknowledge Chinese firms are actively pursuing this capability.

Surveilling the Thought Before You Think It

Leaked internal documents from Geedge Networks, reviewed by Vanderbilt University researchers and reported by The New York Times, show that the firm and its research arm MESA Lab have been developing AI systems designed to identify who could pose a political risk before any act of dissent even occurs.

Geedge already sells a commercial version of the Great Firewall – the software China uses to control and monitor internet traffic. According to the leaked records, the new system pulls together location data, internet activity, and telecommunications records to build behavioral profiles, then runs them through AI models trained to detect what internal documents from February 2024 describe as "harmful information" and to "identify their intent." In China, that phrasing is standard shorthand for political dissent.

Brett V. Benson, a political science professor at Vanderbilt who reviewed the documents, said that Geedge's team was trying to predict what citizens might do next and with whom, with everyday data becoming "raw materials for generating profiles that determine who you are and what you will do next." The infrastructure China has built around state-funded surveillance tools at least pretends to respond to behavior, and this system is designed to get ahead of it.

The Only Thing Standing Between Us and This

The documents suggest Geedge encountered an obstacle to getting their hands on advanced chips. US export controls on graphics processing units limited the firm's access to hardware required for sophisticated AI models, forcing it to rely on older, less powerful systems.

Jimmy Goodrich, a senior fellow at UC's Institute on Global Conflict and Cooperation, put the dependency plainly. Chinese security services face a data overload, and AI's value is in using that data to find threats, but scaling that capability depends entirely on the ability to compute.

That makes chip controls the most consequential privacy protection in this story, which is a genuinely strange sentence to write. The democratic world's leverage over algorithmic pre-crime is an export licensing regime, naturally. Former Biden officials confirmed that those controls have helped maintain the US lead in AI while slowing China's surveillance work.

The Preview Is the Warning

Brett J. Goldstein, director of the Wicked Problems Lab at Vanderbilt's Institute of National Security, described the stakes directly: "This is what happens when mass surveillance meets A.I. Without checks and balances, what China is doing to its own citizens is a preview of what becomes possible anywhere these tools go unchecked."

I think Goldstein is right. Geedge is reportedly still in the research stage, and the system may never be deployed in the form described. And yet, the technical ambition in those leaked records is a commercial product being built by a firm with documented government ties, funded by a state with both the infrastructure and the will to use it. And not in your favor.

The question is what the governments supplying the chips, the cloud infrastructure, and the research partnerships enabling behavioral surveillance at scale think they are moving towards. Authoritarian governments will use this technology the moment it becomes available, and the export controls currently providing friction are the only things those governments have shown any real interest in maintaining.

If you live somewhere that surveils its citizens online, a virtual private network helps protect your internet traffic from that kind of monitoring, and you can get Mysterium VPN for 78% less!


Share on
Facebook share Twitter share Reddit share Linkedin share

Sunscreen, passport, Mysterium VPN. You’re ready.

Get Mysterium VPNArrow icon
general banner img
Dominykas Zukas author photo
Dominykas Zukas
Tech Writer and Security Investigator

Dominykas is a technical writer with a mission to bring you information that will help you in keeping your digital privacy and security protected at all times. If there's knowledge that can help keep you safe online, Dominykas will be there to cover it.

Read more by this author
© Copyright 2026 UAB "MN Intelligence"