AI Surveillance in Authoritarian States for Crime Prevention
Key Findings
Predictive Policing Tools
Predictive policing tools shift state control by turning minor behaviors into lasting records that justify monitoring, allowing authorities to treat nonconformity as risk without requiring illegal acts.
When governments use predictive surveillance systems as part of centralized security, these tools reshape how police and political control interact. They turn early risk judgments into fixed administrative facts. This is clear in how China and Russia use AI scoring in public security. The algorithms deepen bureaucratic habits of stopping dissent before it spreads. Minor behaviors get recorded and turned into lasting digital files. These files justify ongoing surveillance. This allows authorities to treat everyday nonconformity as a threat. No actual crime needs to occur. The result is a steady shift toward constant early intervention. This happens not because technology forces it, but because preserving regime stability outweighs fair process. Most high-risk labels are given without evidence of real criminal danger. Instead of preventing crime, the system builds a scalable way to shape citizen behavior. It matches long-term methods of authoritarian control. Predictive surveillance here acts less to stop crime and more to expand state oversight into daily life. It changes how citizens and the state interact. It does so without changing laws or constitutions.
Predictive Surveillance
Predictive surveillance enforces social control by using AI to detect anomalies in state-controlled data, but only remains effective when unchallenged authority ensures constant data access.
In authoritarian states, centralized surveillance systems use AI to detect unusual behavior. These systems aim to prevent dissent by flagging people for detention or restrictions. They became widespread after 2010, as digital monitoring grew under national security laws in countries like China and Russia. The system relies on the state accessing vast amounts of data without challenge. It assumes no resistance from courts or civil society. This makes predictive scores seem like reliable facts used in governance. The system works best when data flows are stable and state control is strong. If political crises weaken regime unity, data access can break down. Public protests or internal power struggles may disrupt monitoring. When data becomes less reliable, predictive tools lose power. They shift from stopping threats in advance to reacting to current events. Their reach and impact decline as a result.
AI Surveillance Targets
AI surveillance in China targets political threats over crime because the system is built to expand state control, not reduce crime.
AI-powered surveillance is often said to prevent crime by predicting it. This assumes governments mainly want to reduce crime. But in authoritarian states, other goals drive these systems. In China, predictive tools are part of a national security platform. They flag people as high risk based on behavior tied to political dissent or identity. Data shows these systems target ethnic minorities and activists. Arrest and detention records confirm this pattern. The algorithms are not flawed—they are designed this way. They aim to expand state control over society. False alarms are not errors. They are intentional. By casting a wide net, authorities monitor more people. This serves political control, not public safety. Crime rates do not drop. Official reports do not measure crime reduction. The system's real purpose is clear. It stops dissent before it grows. This explains why crime predictions fail. The goal was never crime prevention.
AI Suspicion Machine
Predictive surveillance in authoritarian systems produces widespread social control because its purpose is to expand the number of people subject to intervention, not to improve accuracy.
When governments use AI to predict crime, it changes how much unusual behavior is seen as acceptable. In systems like China's in Xinjiang, facial recognition and behavior tracking flag people based on their ethnicity or religion. These systems rely on hidden risk scores instead of real evidence. The technology is not used to stop crime. It is used to expand state control. Mistaking innocent behavior for risk is favored over missing a possible threat. This creates more people seen as suspects. AI outputs are then used to justify detention or restrictions. The system treats predictions as proof. Over time, this makes mass monitoring and detention routine. It turns statistical guesses into reasons for action. This is not a glitch. It is how the system is meant to work. As a result, entire groups are treated as risky. This pattern is seen in reports from Human Rights Watch and Amnesty International. State data also show rising use of administrative detention. The issue is not bad technology. It is the purpose of the technology. Predictive tools in tight security systems do not improve accuracy. They increase control by design. Expanding the number of people watched or restricted is a feature, not a bug. The system must catch more people to seem effective. So it does.
AI As Control Tool
In authoritarian states, AI-driven surveillance amplifies existing repression because political criteria from party-defined rules, not algorithmic outputs, determine threat responses.
Authoritarian states focus on staying in power above all else. Their systems spot and stop political threats before they grow. This drive exists in security rules, party structures, and old surveillance methods. It predates digital technology, as seen in the Soviet Union, Maoist China, and modern one-party states. Adding AI to public security does not transform social control. It adapts to the existing political logic of centralized threat management. These systems improve how bureaucracies monitor people. But what counts as a threat comes from party rules, not algorithms. Speech, association, and movement are controlled through law and hierarchy. China's Ministry of Public Security and Russia's FSB show this pattern. They have broad discretion under vague legal mandates. So predictive AI tools are used within a top-down framework. Political criteria, not statistical risks, decide when to act. Algorithmic results serve party-aligned security organs, not the other way around. The main mechanism is not AI's autonomy or feedback loops. It is the preexisting need for centralized political control. Any surveillance expansion follows the regime's definition of stability. AI scales existing repression instead of setting new targets.
Predictive Social Control
Predictive surveillance in centralized states replaces reactive punishment with anticipatory discipline through bureaucratic feedback loops, but this mechanism collapses under stress and reverts to overt coercion.
Predictive surveillance systems in centralized states create a feedback loop. They adjust citizen behavior based on anticipated state action instead of punishment. This is seen in China's use of AI risk scores in social credit. The system normalizes compliance through bureaucratic cycles. It shifts the state from reacting to punishing to predicting and disciplining. This structural change blurs law enforcement and social management. The transition fails when predictive accuracy breaks down. This happens during large protests or economic crises. The state then returns to open force. This reveals the limits of algorithmic control and the persistence of traditional authoritarian methods.
