The camera backlash is a trust problem, not a camera problem

A protest against cameras, on the Fourth of July

On a holiday about independence, a group of residents in Lubbock, Texas showed up to protest automated license-plate readers. Their demand was simple: take the Flock Safety cameras down. Think about what it takes for people to spend the Fourth of July organizing against a security camera. That is not a technology complaint. That is a trust collapse.

Lubbock is not alone. Across the country the mood has turned hard against these cameras: dozens of cities have moved to cancel or reject their contracts, systems that run billions of plate scans a month, and in at least one town people physically cut the cameras down off their poles.

It is tempting to read all of this as people rejecting the technology. I do not think that is what is happening. They are rejecting how it arrived: quietly, everywhere at once, with no clear answer to the only questions that actually matter. One resident put it bluntly in an online thread about the Lubbock cameras: "How do you find out who approved it? Who owns it? Wtf happened to privacy?" Those three questions are the whole story.

The tell is in what one city argues about

Lubbock has run these cameras for about a year, and listen to how the two sides describe the exact same hardware.

The police department: "Flock cameras are not surveillance. They are an investigative tool. We use them on a daily basis to help solve cases." A local defense attorney: "This isn't police work. This is a 24-hour surveillance system, a dragnet that has been placed over the city. Everybody, not suspects, not criminals, every person who comes in or leaves is having their picture taken."

Both are describing the same cameras on the same poles. The fight is not about whether the technology works. It is about what gets collected, how long it is kept, where it travels, and who is allowed to point it at whom.

"Where does it go" is the question nobody answered first

On paper the retention sounds restrained. Images not tied to an active case are deleted after 30 days. That is the answer that gets offered when someone raises privacy.

But the data does not just sit in Lubbock. These systems are built to share across a nationwide network, and the Electronic Frontier Foundation found that a single local search can fan out across thousands of agencies at once. In one documented case, Lubbock police ran a search with the logged reason "protest veh" that reached nearly 6,000 networks. In other places, officers justified a search by typing a single word: "protest."

The ACLU put the underlying problem plainly: "A lack of transparency, oversight, and regulation into how they collect, store, and use our data, and how to hold public and private actors accountable if they abuse it." That is the entire distance between "an investigative tool" and "a dragnet." It was never the camera. It was the collection, the retention, the quiet nationwide sharing, and the near-total absence of anyone auditing it.

Safety tech does not lose people because it can detect. It loses them when it collects everything first and explains itself later.

The questions that decide whether you are trusted

If you build or buy anything that watches a public or shared space, these are the questions that separate a tool people accept from a system they organize against:

• What do you actually collect, and is it the minimum the job needs?
• Where does it physically live, and does it ever leave the building?
• Who owns it, sets the retention, and is allowed to query it?
• Is it scoped to a stated purpose, or can it quietly be pointed at anyone?
• Can any of that be audited by someone who is not the operator?

None of these are about capability. A system can be technically excellent and still fail every one of them. That is exactly how you end up with a Fourth of July protest instead of a safer city.

How we think about it

This is the reason DHI is built the way it is. Detection runs on-site, on the customer's own hardware. Raw video never leaves the building, so there is no nationwide pool to share into and no vendor cloud holding your footage. The customer owns the retention policy and the purpose, and the system is scoped to a specific safety job rather than open-ended monitoring of everyone who passes by.

It is the same point we made about facial recognition: the capability is not what makes something surveillance. The architecture is. Collect less, keep it local, scope it to a purpose, and be able to explain all of it on day one.

Trust is the feature

The companies whose cameras are getting cut down did not lose because their technology was weak. They lost because they treated trust as something to manage after deployment instead of something to design in before it.

For anything that watches people, trust is not a compliance checkbox. It is the product. Build for it, and you are still deployed when the backlash clears. Skip it, and it does not matter how good the detection was.

Sources: Chron (July 4, 2026) on the Lubbock protest, KCBD (June 4, 2026) on the local camera debate, the EFF's analysis (November 2025) of nationwide ALPR search sharing, and the Washington Times (July 2, 2026) on the national backlash.