Fifty Officers Isn't a Personnel Problem. It's a Product Problem.

The headline calls them "rogue officers."

That framing deserves some scrutiny.

According to Washington Post reporting by Drew Harwell and colleagues, at least 50 law enforcement officers have been charged with or accused of misusing Flock Safety's license-plate camera network — and other similar systems — for purposes that include stalking. More than half of the documented cases involved surveillance of romantic partners. Fifty officers. Multiple agencies. A pattern distributed across the country.

One rogue officer is a personnel problem. Fifty is a design problem.

The distinction matters because the two diagnoses lead to completely different responses. A personnel problem gets solved with terminations, retraining, and strongly worded policies. A design problem requires architectural changes to the product itself — access controls, mandatory audit logs, use-case restrictions, automatic alerts when a query pattern looks like a personal investigation rather than a law enforcement one. Personnel responses are cheap and visible. Product redesigns are expensive and largely invisible to the public.

The surveillance industry has every incentive to accept the personnel diagnosis.

Here is how Flock Safety's system ordinarily works, as far as the public record shows: cameras mounted on roads and in parking lots capture license plates and vehicle characteristics. That data flows into a searchable database. Officers query the system by plate number and receive location data — where the vehicle was seen, when, and in what direction it was traveling. One query is a data point. Repeated queries on the same plate, across different times and locations, become a map of someone's life.

The product is extraordinarily powerful. That is the point. Flock Safety has marketed it aggressively to local governments and law enforcement agencies as a crime-fighting tool, and the company has raised hundreds of millions of dollars in private investment on the strength of that pitch. Investors do not fund surveillance infrastructure at that scale because they enjoy civic improvement. They fund it because the data asset appreciates as the network grows.

Which is exactly why the access control question matters.

A system designed with genuine use-restriction in mind would require officers to log a case number before running a plate. It would flag repeated queries on a single plate originating from one officer's login. It would generate automatic supervisor notifications when query patterns deviate from normal investigative behavior. It would limit data retention windows that allow historical movement reconstruction. These are not exotic technical requirements. They are standard governance features in any enterprise software system where the data is sensitive and the abuse potential is high.

The question the "rogue officers" framing never asks is: were any of those features present?

Fifty cases of misuse suggest they were not — or that they existed on paper and generated no real consequences in practice. One bad actor can defeat a well-designed system. Fifty bad actors in multiple agencies, apparently over a sustained period, cannot defeat a well-designed system. They can only defeat a system that was not designed to stop them.

Now, the strongest counterargument deserves an honest hearing. A skeptical reader might note that 50 cases drawn from a network deployed to thousands of agencies could represent a small fraction of total users — and that every technology capable of misuse will eventually find someone willing to misuse it. Firearms are misused. Police databases are misused. That does not make every tool a defective product. The proper response, on this view, is exactly what the "rogue officers" frame suggests: prosecute the individuals, improve training, and move on.

That argument holds up until you ask what the prosecution rate implies about detection. Fifty charged or accused cases means fifty instances that surfaced — through victim complaints, whistleblowers, or supervisors who happened to notice. Misconduct that results in charges is almost always a fraction of total misconduct, because charges require detection, and detection requires that the system generate evidence of misuse. If the system does not audit itself, most misuse is invisible. Fifty visible cases in a network of this scale is not reassuring. It is the visible edge of an unquantified problem.

To the company's partial credit, Flock Safety did not stay silent. After the Post's reporting, it announced changes to its system and its CEO publicly acknowledged the misuse. That response is the right direction. It is also a tacit admission of the argument here: if the answer to fifty stalking cases is to change the product, then the product — not just the officers — was the problem. The open questions remain the concrete ones: what logging exists, who can see it, and what actually triggers a review. Those are exactly the factual questions that should be part of any public discussion about renewing or expanding these contracts.

Cities and counties across the country are currently deciding whether to sign or extend agreements with Flock Safety and its competitors. Those decisions are being made at the city council and county commission level, often with minimal public attention. The standard pitch is crime reduction. The contract terms governing data retention, access logs, third-party sharing, and audit rights are rarely the subject of the conversation.

They should be the entire conversation.

The judge in any individual stalking case will decide what happens to the officer. The public — through elected officials — decides whether the underlying product gets to keep operating without the controls that would make the next fifty cases detectable before they become fifty more.

Start with the audit logs. Find out if they exist. Then ask who has ever looked at them.