How large is the system?

The police-tech giant Flock operates a network of some 120,000 automatic license plate readers around the US. The company announced a set of changes to its platform meant to prevent officers from using it for illegal or illegitimate purposes.

The reason for the changes is concrete. The Washington Post identified 50 cases in which systems from Flock and its competitors were misused, often to stalk and harass women.

The documented cases

Two examples reported by MIT Technology Review show the nature of the problem:

  • In Wisconsin, a woman alleged that her officer ex-boyfriend searched for her car 179 times.
  • Another woman was being stalked by the chief of police, with nobody to report him to.

The second example points to a problem no technical safeguard can solve: when the authority meant to oversee the system and the person misusing it are the same, internal review does not function.

The measures introduced

Flock's response consists of practices aimed at ensuring officers have a proper cause for every search. Two stand out: using software to flag abnormal searches, and requiring searchers to enter a criminal case number.

On paper this ties every search to an open investigation. An audit trail forms, each query matches a case, and unusual behaviour becomes statistically detectable.

Where the loopholes are

The changes come with big loopholes. The most obvious sits in the case-number requirement itself: officers can enter bogus case numbers. This is not a hypothetical concern — officers have lied to get around other Flock safeguards before.

The structural problem is this: the check relies on a declaration by the person being checked. The system asks which case a search is for, but does not verify the answer. Verification would require matching the entered number against a real case, which means connecting not to Flock but to each individual department's records system.

The second gap is broader: the new policies do not address the more fundamental issues arising from the system's existence. Who can be searched under what conditions, how long records are kept, and who has access to the network all remain separate open questions.

Why it matters

This debate shows a pattern that recurs with AI-assisted surveillance tools. The system is built for a legitimate purpose: finding stolen vehicles, catching wanted drivers. But the capability that emerges is not bounded by that purpose; a network of 120,000 cameras can describe retroactively where a car has been.

Misuse is therefore not an exception but a predictable outcome. Technical measures help but are not sufficient; what decides the result is who holds access and whether oversight comes from outside the system. That the 179 searches in the Wisconsin case went unnoticed for so long makes the point: the records existed, but nobody was looking.