What it is

Sensing that is everywhere and always on: cameras in streets and shops, microphones in devices, location from phones, purchases, messages, and records. On top of the sensing sits inference: systems that conclude not only what a person did, but what they feel, intend, fear, and could be moved by.

A small example: a shop’s cameras recognize a returning customer, a model guesses from her face and pace that she is stressed, and the prices and offers she sees on her phone change to match.

What changes

For What changes
A person Privacy stops being the default. Knowing they may be read, people say less, try less, and conform more, which erodes interior sovereignty.
A town or region Public space becomes monitored space; protests, gatherings, and ordinary life are recorded and analyzed.
A firm Data becomes a source of advantage, and inference a product to sell.
A government The means to watch everyone becomes cheap; the temptation to use it in a crisis becomes hard to resist.

Signs it is becoming ordinary

Sign What it indicates
Inference about mood, intent, or belief sold or used in decisions inference has become a market
Facial recognition built into public space by default sensing is ambient, not occasional
Records joined across agencies and firms the firebreaks are gone
Data brokers holding detailed profiles on most adults the raw material already exists
People changing behavior because they assume they are watched the chilling effect has arrived

The readiness pair

Must exist first: limits on what may be inferred about a person, firebreaks that keep data from being joined across domains, and the right not to be read.

Its components:

  1. Limits on inference: some conclusions may not be drawn or used at all, such as emotional state, belief, or health, in decisions about people.
  2. Firebreaks: data gathered for one purpose may not be joined with data gathered for another without fresh, public justification.
  3. The right not to be read: public spaces where recognition is off by default, and the right to refuse sensing in private ones.
  4. Deletion that works: a person can have data about them erased across brokers in one request.
  5. Use limits in a crisis: emergency powers to watch expire on a fixed date and never touch interior life (coercion does not acquire the person).
State What it looks like
Started prohibitions written on certain kinds of sensing; deletion rights created
In place prohibitions enforced; deletion requests processed across brokers
Tested a large deployment of prohibited sensing stopped by a court or regulator; inference in a decision found and barred

Why it cannot wait

Data once gathered and joined cannot be ungathered. Every year without firebreaks adds records that will outlive any later rule. Limits on collection, which most rules address, do not reach inference, which can be drawn from data already held. Readiness has to cover what may be concluded, not only what may be collected.

When readiness comes late

A city installs recognition cameras to find serious offenders, with promises that the footage will be used for nothing else. Years later, the system’s records are joined with transit and payment data to find who attended a protest. Nothing new was collected; the inference was drawn from what was already there. People stop attending. That is AI surveillance: watching that becomes self-policing.

What to start now

Who Step
A person use deletion rights; refuse sensing where possible; keep some life off networked systems
A town keep recognition off by default in public space; publish every sensing system and its purpose
A firm refuse to draw or sell inferences about mood, belief, or health; keep data sources separate
A government write limits on inference, not only collection; build firebreaks between agencies’ records; make emergency powers expire

What would show it working

Inferences about belief and emotion absent from decisions; records that stay in their own domains; deletion requests honored across brokers; and people behaving in public as if they were not being read, because they are not.