Surveillance Anxiety: When Safety Was Not the Real Reason

Matthew Sexton, LCSW, NATC — Licensed clinical social worker in private practice, providing telehealth psychotherapy to adults in New York.

Most people do not walk in describing themselves as anxious about surveillance. They describe an ordinary drive that turned into something they were thinking about, a pole they now notice on a road they have taken for years, and then a quieter worry underneath it: that noticing at all makes them strange.

It does not. That confusion is the honest starting place, and it deserves better than being talked out of. What follows is the documented version of what happened, the clinical name for the particular injury involved, and the reason a body can keep responding to it long after the news cycle moves on.

Sold as safety, logged as something else

The Texas search logged the reason "had an abortion, search for female" and reached 83,345 cameras. A system sold for one purpose recorded another, and the record is what makes this checkable rather than a matter of opinion.

Automatic license-plate reader networks arrived in communities as a safety tool. The pitch is familiar: recover stolen cars, find missing people, close cases faster. At least 80,000 of these cameras now sit across the United States (TechCrunch, 2026). Flock, the company behind the largest network, says the system logs roughly 20 billion plate scans a month across 49 states. That last number is a company claim, and no independent party has audited it (Common Dreams, 2026).

The raw count matters less than the distance between a stated justification and a recorded one.

In Johnson County, Texas, a sheriff's official ran two searches of the Flock network. The reason string logged for them read "had an abortion, search for female." The first covered 1,295 networks and 17,684 cameras. The second covered 6,809 networks and 83,345 cameras, reaching into states where abortion is legal. The Electronic Frontier Foundation documented both from search logs, incident reports, and detective affidavits (EFF, 2025).

Sheriff Adam King told reporters it was a welfare check. Her family, he said, "was worried that she was going to bleed to death." Court records describe something else. A detective's sworn affidavit refers to a "death investigation" of a "non-viable fetus," documents the collection of abortion records, and notes a consultation with the district attorney about whether charges could be filed. The case stayed open more than a month (404 Media, 2025).

Two independent reporting efforts, working from different material, arrived at the same fact. The reason given to the public and the reason typed into the search field were not the same reason.

The problem here is structural rather than personal. A far-reaching system was presented to communities as protection, and its own use log recorded a different purpose.

That difference changes the question you are sitting with. It is no longer whether cameras are good. It is whether the stated boundaries of a system tell you what the system will do.

Why that gap is the actual injury

Because it turns ordinary surveillance into institutional betrayal, which research treats as a distinct injury rather than a worse version of the same one.

Carly Smith and Jennifer Freyd defined institutional betrayal as wrongdoing perpetrated by an institution upon people who depend on that institution, including a failure to prevent harm or to respond honestly once harm occurs inside the institution's context (Smith & Freyd, 2014).

Being watched is one category of experience. Being watched under an explicit promise of safety, and later learning the promise was not the operative reason, is a different injury. That second one is what this article is naming.

The construct carries real weight outside of cameras. Earlier work in the same research program found that institutional betrayal compounded outcomes among women who had experienced sexual assault, accounting for variance in anxiety, dissociation, and sexual dysfunction beyond what the original event predicted (Smith & Freyd, 2013). Later scholarship extended the framework to state actors.

None of that transfers directly to license-plate readers, and this article is not claiming it does. What transfers is narrower and still useful: reliance is what makes institutional wrongdoing psychologically distinct. The betrayal is not that cameras exist. It is that a dependency relationship got used in a way the public framing never disclosed.

If you have been quietly asking whether you are overreacting to something that is supposed to feel fine, that question belongs here. It shows up in every high-control arrangement, and it is usually tracking something real.

The pictures expire. The record of who looked does not.

One structural detail makes the logged-reason problem possible, and it comes from a primary court document.

Flock's own policy states that license-plate data is "hard deleted on a rolling 30-day basis by default," and that the schedule may be increased or decreased when a customer's law or policy requires it. Thirty days is a default, not a ceiling (Flock Safety LPR Policy, updated 2026).

Sworn testimony recited by the Washington Court of Appeals goes further. Image data purges at thirty days, but "An audit log serves as a record of every database inquiry. The audit log is maintained indefinitely." Every query also requires the user to specify a reason (State v. Simonson, 2026).

So the photograph of a car passing through town rolls off. The record of who went looking, when, and under what stated reason does not. Those Texas reason strings exist precisely because of this logging.

The scale is not imaginary either. EFF obtained datasets covering more than 12 million searches by more than 3,900 agencies between December 2024 and October 2025, a records-based figure rather than a marketing one (EFF, 2025).

You do not need to believe anyone is following you personally to react to that arrangement. Part of the weight of an ambient system is never knowing when it becomes relevant to your own life.

The pushback is already happening

You do not have to become an activist to notice that institutions themselves are split on this.

The Los Angeles Police Department, the third largest in the country, let its three-year Flock contract expire. Its CIO, Dean Gialamas, put the reason on record: "This contract is not being renewed because of serious concerns around civil liberties and civil rights issues, particularly around privacy and the data that is being collected from these cameras" (TechCrunch, 2026).

In Dayton, Ohio, a city audit found Flock data searched for immigration purposes more than 7,000 times by outside entities, in violation of the city's own policy. Workers covered 72 fixed cameras with trash bags (Fortune, 2026). California's Attorney General sued the City of El Cajon over sharing plate data with 27 out-of-state agencies (California Office of the Attorney General, 2025).

Florida's verified record follows its own pattern, and it is worth stating precisely. Five Treasure Coast agencies ran 216 immigration-related Flock searches between January and March 2026, against zero in the same window a year earlier (TCPalm via AOL, 2026). In July 2026, the Leon County Commission voted to strip Flock funding out of a grant package after hours of public opposition (Tallahassee Reports, 2026). Nothing in Florida's documented record involves an abortion search. That fact pattern belongs to Texas, and the two should not be blurred together.

Named agencies, named dates, opposite local choices. Your unease is not occurring in a vacuum.

Surveillance anxiety runs on uncontrollability, not visibility

Because the operative variable is uncontrollability, not visibility. A stressor you cannot end is measurably harder on a body than the same stressor with an off switch.

The stress research that matters most here studies control rather than cameras. Glass and Singer exposed people to bursts of loud, unpredictable noise. Participants who held a button that could stop the noise showed less autonomic reactivity, better performance afterward, and smaller lingering effects. Most of them never pressed it. The mere availability of control, not its use, changed the stress response, and the uncontrollable plus unpredictable condition produced the worst outcomes (Glass & Singer, 1972).

That finding got summarized badly for decades, as though people simply learn to be helpless. Neuroscience reversed the account. Passivity under a prolonged aversive event is the default, unlearned response. What actually gets learned is control, through a prefrontal circuit that detects controllability and inhibits that default (Maier & Seligman, 2016).

That correction matters for how you read your own fatigue. Feeling worn down by this is not a failure to stay strong. What an ambient system takes is the opportunity to learn you have any say at all.

Observation alone changes behavior, too, without any consequence landing. Preschoolers told a television camera would record them showed measurably less interest in the same activity two weeks later, judged by unobtrusive classroom observation (Lepper & Greene, 1975). Being watched turned play into work.

Adult evidence points the same way. A meta-analysis of 94 samples covering 23,461 workers found electronic performance monitoring associated with increased stress regardless of the characteristics of the monitoring, with no evidence it improved performance. Transparent, less invasive monitoring produced better attitudes than covert, invasive forms (Ravid et al., 2023).

What the mechanism predicts for a system that already runs hot

No research studies automated license-plate networks and trauma. What follows is a mechanism argument, and it deserves that label every time it appears.

Allostatic load describes the wear a stress response leaves behind when it runs longer than it should. One of its four documented conditions is a prolonged response caused by delayed shut-down (McEwen, 1998).

A trauma-shaped threat-monitoring system already struggles with exactly that. Research on attention in PTSD describes biased and unstable deployment: attention swings toward danger cues and then away from them, matching both the hypervigilance and the avoidance side of the picture. Greater variability in that attention distinguishes PTSD from healthy comparison groups (Naim et al.).

Put the pieces beside each other without overclaiming. Ambient plate readers run continuously. You did not opt in. You cannot query the log, because you are not the party holding it. Public justifications and recorded reasons can diverge, which removes the "I know what this is for" anchor. The mechanism predicts that a monitoring system already prone to delayed shut-down will treat that combination as input it cannot finish. Nothing in the environment returns a trustworthy done signal, so the assessment restarts.

That is a prediction about a mechanism, not evidence that any camera has harmed anyone in particular, and not a diagnosis of you.

The nearest large study measured something adjacent. Across 8,066 people in 34 New York City neighborhoods, researchers linked neighborhood-level stop-and-frisk and use-of-force rates to higher nonspecific psychological distress among men. The same study found no comparable association among women (Sewell et al., 2016). Cite it for what it actually is: evidence that living under visible enforcement surveillance can register in population mental-health data, with a gendered result worth reporting accurately rather than flattening.

Behavior shifts as well, modestly. Traffic to privacy-sensitive Wikipedia articles fell roughly 30% after the June 2013 Snowden disclosures, including a change in the longer-term trend rather than a brief dip (Penney, 2016). The field's own summary is that chilling effects are real and often small. Several methods lean the same direction, which is enough to take seriously and not enough to claim surveillance silences everyone.

If your attention has felt expensive lately, what therapists actually mean by nervous system regulation covers the same body mechanics without the news attached.

What helps with surveillance anxiety

Name the breach precisely, sort what you know from what you do not, locate the control that is still yours, and schedule recovery rather than waiting for reassurance. Each of those is doing specific work, described below.

Naming the experience precisely is itself a regulating move. Diffuse dread that acquires a specific name tends to get smaller, because your system stops treating everything as the problem. You are allowed to say: I am reacting to a mismatch between a safety promise and a documented use, and my body is doing what bodies do when shut-down never arrives.

Then sort it into three columns, on paper or in your head.

What is known. Keep it short and factual. A documented search carried one reason while the public got another. The system is large. Your concern has a real object.

What is unknown. You do not know who will search what, or when a policy will change. Naming that as uncertainty keeps it from quietly hardening into certainty about the worst outcome.

What control is available. This is the column that does the work. You do not control the network. You do control when you read about it, whether you are reading contract fights at midnight, who you talk to when the dread spikes, and whether you keep the appointments you need. That is not a consolation prize. In the research above, availability of control is the active ingredient.

Then schedule the shut-down rather than waiting for reassurance that is not coming. Pick a point in the day where you stop reading and stop scanning, and follow it with something repeatable: a walk, cooking, music, a conversation that is not a briefing on the news. You are teaching your attention to leave and come back by choice.

Your reaction is proportionate to a real documented pattern. That sentence is worth reading twice if you have been minimizing it. The habit of asking whether something was bad enough to count shows up here like it shows up everywhere else. You do not need a personal encounter with a search log to register a broken promise at the system level.

Last, watch for withdrawal. Institutional mistrust carries a measurable downstream cost: people pull back from services they need, a pattern documented around prevention and emergency care (Harm Reduction Journal, 2022). Notice if you are delaying a medical visit, skipping something useful, or isolating where you would normally reach out. That pullback is not proof you are well informed. It is usually the cost landing.

None of this requires talking yourself into trusting a system that has not earned it. The work is keeping justified mistrust from taking over parts of your life that had nothing to do with it, which is the slower business of taking back your agency.

If you want help sorting the body's part of this from the news cycle's part, that is work I do with adults over telehealth. Book a call when you are ready. You do not need a cleaner story first.