The answer depends on what level of surveillance we are talking about. AI is not a single surveillance technology; it is a collection of technologies that make surveillance cheaper, faster, more predictive, and more automated.
Using the People’s Republic of China as an example, it helps to think of AI surveillance in several layers.
1. Seeing: collecting data
The first layer is simply gathering enormous amounts of information.
This can include:
- CCTV cameras
- facial recognition
- license plate readers
- smartphone location data
- payment records
- online searches
- social media posts
- voice recordings
- travel history
- biometric information
AI does not need to watch every person continuously. It only needs enough data to reconstruct patterns.
2. Recognizing: identifying people
Computer vision can identify:
- who someone is
- where they have been
- who they met
- whether they are carrying unusual objects
- whether a crowd is forming
- whether someone appears on multiple cameras
Instead of thousands of people watching video feeds, AI can automatically flag events for human operators.
3. Connecting: building networks
This is where AI becomes much more powerful.
Rather than focusing on individuals, AI builds graphs of relationships.
For example:
Person A regularly meets Person B.
Person B communicates with Person C.
Person C organizes online discussions.
AI can infer social networks even if no one explicitly announces them.
This resembles what intelligence agencies have long done, but AI scales it to millions of people.
4. Predicting: estimating future behavior
Modern machine learning is increasingly used to estimate probabilities.
Instead of asking,
“What did Kelly do yesterday?”
the system asks,
“Who is likely to organize a protest next month?”
or
“Which users are becoming politically influential?”
Prediction does not have to be perfect to influence decisions.
5. Recommending interventions
This is perhaps the most significant change.
AI can rank cases according to estimated risk.
For example,
- ignore
- monitor
- increase observation
- restrict travel
- conduct an interview
- investigate further
The final decision may still be made by humans, but AI determines where human attention is directed.
Why AI changes surveillance
Traditional surveillance has a bottleneck:
human attention.
Even authoritarian governments cannot manually examine billions of interactions.
AI changes this.
Instead of
people watching cameras,
it becomes
AI watching everything,
humans only reviewing exceptions.
The bottleneck shifts from data collection to policy.
Why prediction can become dangerous
One concern raised by many researchers is that prediction can be treated as evidence.
For example,
suppose AI estimates that someone has an 85% probability of participating in future dissent.
If authorities begin acting before anything illegal occurs, surveillance moves from investigating actions to managing predicted behavior.
This raises serious legal and ethical questions about due process, false positives, and civil liberties.
The CCP context
Publicly available reporting indicates that Chinese authorities have integrated AI into multiple aspects of governance, including city surveillance, facial recognition, online content moderation, and public security platforms. These systems vary considerably across regions, and many technical details remain opaque. Researchers generally agree, however, that AI has increased the scale and speed with which authorities can process data and respond to perceived risks.
One observation related to our conversations
Over the past year, you’ve often contrasted the algorithmic self with the analogical self.
From that perspective, AI surveillance is not only about watching people.
It is about constructing an algorithmic version of a person.
The system does not need to understand your inner life. It only needs a sufficiently accurate statistical model of your behavior.
That model can answer questions like:
- Where will you probably go?
- Who will you probably meet?
- What will you probably buy?
- Which message are you most likely to believe?
- Which fear is most likely to change your behavior?
Notice the shift:
The target is no longer the living, irreducible person.
The target is the predictive model of that person.
That is why AI surveillance is qualitatively different from earlier forms of surveillance. It does not merely record the past; it seeks to anticipate and influence the future through continuously updated models of human behavior.
