A Government That Wants to Predict Who Its Enemies Will Be
China has spent years building one of the world’s most sophisticated surveillance systems. Cameras track movement, internet controls shape what citizens can see, and online speech is heavily monitored. Now, according to reporting from The New York Times and researchers studying leaked documents, China appears to be taking the next step: using artificial intelligence not only to watch dissidents, but to predict who might become one.
The idea sounds like science fiction. Yet according to a June 2026 New York Times report by Julian E. Barnes, a Chinese company has been developing AI-powered tools designed to identify people who could someday pose a political risk to the government. The technology would not simply look for critics who already exist. It would try to forecast dissent before it happens.
The Times described the effort as a glimpse into a future in which “an authoritarian state is able to move against its citizens before they begin any public dissent.” In other words, China appears to be moving toward a system that does not wait for criticism. It searches for warning signs and acts first.
The Company Building Political Prediction Software
At the center of the report is a Chinese company called Geedge Networks. Geedge already sells what the Times described as a commercial version of China’s Great Firewall, the massive surveillance and censorship system the Chinese Communist Party uses to regulate online life.
The Great Firewall is more than a filter. It is a system of control. It monitors internet traffic, blocks forbidden websites, suppresses politically sensitive topics, and flags attempts to evade censorship. Citizens cannot simply browse freely. Their access to information is shaped by what authorities allow them to see.
According to leaked company documents reviewed by Vanderbilt University researchers, Geedge is now working on AI tools that go much further. The company, along with its government-supported research arm MESA Lab, has reportedly been developing software that generates profiles of citizens and identifies who may present a political risk.
That raises a chilling question: What happens when a government decides that disagreement itself is dangerous?
According to the Times, Geedge researchers were building behavioral profiles using telecommunications data, social media activity, and location information. The software would look for patterns in daily life, searching for clues about what someone believes, where they go, who they associate with, and what they may do next.
Researchers found discussions inside the company about connecting physical movement to online activity, including what books people read and what movies they watch. In one February 2024 meeting, employees reportedly discussed how to build systems to “identify their intent” and “achieve discovery of harmful information.”
That phrase matters because “harmful information” is often used by Chinese authorities as a euphemism for political dissent or material the Communist Party wants suppressed.
The technology would not merely search for forbidden posts or political slogans. It would attempt to infer future behavior.
Brett V. Benson, a political science professor at Vanderbilt University who examined the leaked documents, said Geedge researchers were “doing more than just documenting behavioral patterns.” Instead, he explained, “They were trying to predict what citizens might do next and with whom.”
In effect, ordinary behavior becomes evidence. A location history, a reading habit, an online discussion, or even an association with certain people could become part of a digital political profile.
When Surveillance Meets Artificial Intelligence
Critics say the danger grows dramatically once surveillance systems are combined with artificial intelligence.
“This is what happens when mass surveillance meets A.I.,” said Brett J. Goldstein, director of the Wicked Problems Lab at Vanderbilt’s Institute of National Security.
Goldstein warned that what China is doing should concern people far beyond Beijing. “Without checks and balances,” he said, “what China is doing to its own citizens is a preview of what becomes possible anywhere these tools go unchecked.”
That warning is especially alarming because AI changes the scale of repression.
Traditional surveillance requires armies of police, censors, analysts, and informants. AI can automate much of the process. Instead of human workers combing through millions of conversations, algorithms can sort through massive amounts of information in seconds. Instead of monitoring one protest, software can continuously scan an entire society.
Freedom House, in its report The Repressive Power of Artificial Intelligence, warns that AI acts as “an amplifier of digital repression,” making censorship, surveillance, and propaganda easier, faster, cheaper, and more effective.
That may be the most important takeaway. AI does not invent authoritarianism. It supercharges it.
How AI Can Become a Tool of Repression
The reports describe several ways governments can use AI to tighten political control.
Mass surveillance is perhaps the most obvious. AI systems can rapidly process huge amounts of telecommunications data, online activity, location tracking, and digital behavior. Instead of simply collecting information, authorities can use algorithms to rank risks and prioritize targets.
Facial recognition technology takes surveillance into physical space. The Journal of Democracy reports that China has built vast networks of AI-driven cameras capable of identifying people in real time. Public gatherings, protests, and even ordinary daily activities become visible to authorities.
Predictive policing adds another layer. AI systems can analyze surveillance footage, social media activity, police records, and databases to forecast future unrest or criminal behavior. The Journal of Democracy warns that such systems can unfairly label activists or dissidents as threats before they have done anything illegal.
China has also used AI to target the Uyghur community under what critics describe as the guise of counterterrorism. The technology has reportedly been used to monitor populations, identify individuals, and reinforce political control.
Automated censorship makes speech control far more efficient. Freedom House says governments increasingly require platforms to use machine learning tools to detect and remove political, social, or religious content at speeds impossible for human moderators alone.
The effect is subtle but powerful. Instead of a visible censor blocking speech, algorithms quietly erase it.
AI also helps spread propaganda and disinformation. Another Chinese company discussed by researchers, GoLaxy, reportedly worked on software capable of pushing targeted propaganda that supported Beijing’s preferred narratives while opposing ideas the government disliked.
Even chatbots can become instruments of control. Freedom House reports that Chinese generative AI systems are required to follow “core socialist values” and exclude politically undesirable material. Some Chinese chatbots reportedly refuse to answer questions about subjects such as Tiananmen Square while repeating official Communist Party narratives about Taiwan.
In such an environment, information itself becomes curated reality.
The Export of Authoritarian Technology
What may worry critics most is that China’s model may not remain inside China.
The Times reported that Geedge has exported network security software to countries including Ethiopia, Kazakhstan, Myanmar, and Pakistan, helping enable mass surveillance on mobile networks.
If predictive political surveillance becomes part of that export package, China may be creating something larger than a domestic surveillance state. It could become a supplier of authoritarian technology for governments seeking stronger control over citizens.
That possibility fits a broader warning raised by democracy advocates. Freedom House argues that AI-powered repression is spreading because it lowers costs, increases efficiency, and reduces the visibility of censorship and surveillance.
A government no longer needs tens of thousands of people policing speech when algorithms can do much of the work.
The Fear May Matter More Than Accuracy
One of the most unsettling observations came not from officials or researchers, but from a reader responding to the Times story.
“The real problem here,” one commenter wrote, “is that the technology doesn’t even have to be correct in identifying the real dissidents. It only has to exist and be known to the public to be an effective suppressant on free speech.”
Julian Barnes, the Times reporter, responded simply: “Exactly right.”
That insight may reveal the deepest danger.
A predictive system does not need to be perfect to reshape behavior. If citizens believe their searches, friendships, travel patterns, reading habits, or conversations are constantly analyzed for political risk, many may silence themselves before authorities ever intervene.
Fear becomes self-policing.
China’s critics see a country moving steadily toward a future once imagined only in dystopian fiction, a society where algorithms do not merely punish dissent, but predict it. A place where surveillance is constant, information is filtered, speech is monitored, and political disagreement becomes something identified and managed before it can grow.
The technology is still developing. But the direction is already becoming clear.