Digital surveillance promises to make institutions more secure by enabling them to see more. Automated license-plate readers, facial recognition systems, and interconnected databases allow authorities to identify patterns and retrieve information at speeds once considered impossible. Yet greater capacity to detect and classify does not necessarily improve the ability to determine what happened. An institution may acquire more information without gaining the ability to interpret it accurately.
The distinction matters because surveillance does more than collect information. Its outputs can initiate investigations, influence decisions, and authorize interventions that affect people who have little control over how the information is interpreted. The central problem is not merely that surveillance systems sometimes make mistakes. Rather, an institution can transform an uncertain indication into an apparently established fact through the very procedures the indication sets in motion.
Recent cases involving automated license-plate readers illustrate this problem. In “The High Cost of the Automated Flock-Camera Surveillance State,” published by the Mises Institute, Ryan McMaken documents instances in which police officers used Flock data to track people for personal reasons. He also describes a Colorado case in which an officer accused a woman of theft after a false match and refused to consider evidence of her innocence. In another incident, a mother and her children were held at gunpoint after their vehicle was incorrectly flagged as stolen. These cases involve different circumstances and do not establish that automated license-plate readers are inherently unreliable. They do, however, raise a common question: what happens when information generated by a technical system acquires practical authority before its accuracy and relevance have been adequately established?
The crucial distinction is between detecting an indication and establishing a fact. A camera may record a vehicle at a particular location or suggest a possible link to a reported offense. By itself, it cannot establish who was driving, why the vehicle was present, or whether an offense occurred. Those questions require independent verification and contextual judgment. An automated match can serve as a useful investigative lead without amounting to proof of wrongdoing.
The institutional risk begins when the line between a lead and proof blurs. An alert prompts an investigation; the investigation produces reports and records; and those records may later be treated as confirmation that the original suspicion was justified. Each step can appear procedurally legitimate without resolving the uncertainty that existed at the outset. The institution need not fabricate evidence or act in bad faith. It may instead mistake the procedural development of a case for independent confirmation of its underlying premise. An uncertain signal thereby acquires authority through the very process it initiated.
This is a problem of institutional reasoning as much as of technological accuracy. The initial error need not be repeated in the same form. It can persist through the assumptions embedded in later decisions. Once a person or vehicle is associated with suspicion, subsequent actions may be interpreted through that lens. Information that supports the original assumption receives attention, while contradictory information may be discounted or ignored. If the resulting records remain accessible, the initial suspicion can influence future decisions even after its evidentiary basis has weakened.
The Colorado case reported by McMaken underscores the importance of this distinction. The relevant question is not merely whether the system produced a false match, it is whether the institution adequately tested that match against contrary evidence before acting on it. A technological error becomes an institutional failure when procedures fail to correct it, and decision-makers continue to treat its consequences as grounds for maintaining the original suspicion. The available account illustrates this danger, though it cannot establish how often the same sequence occurs across police departments.
Friedrich Hayek’s analysis of dispersed knowledge helps clarify the broader problem. In “The Use of Knowledge in Society” (1945), Hayek argued that socially-relevant knowledge is distributed among individuals and cannot be assembled into a complete body of information available to a central decision-maker. Applying this insight to digital surveillance does not mean Hayek directly analyzed contemporary camera networks. Rather, it shows that his argument identifies a limitation relevant to them: extensive data do not eliminate the need for contextual knowledge that the data may not contain.
A searchable database can pinpoint where a vehicle was recorded and identify possible connections among records. It cannot automatically provide all the circumstances needed to interpret those connections. The system may make information more legible to an institution while leaving the underlying event incompletely understood. The danger arises when increased legibility is mistaken for increased certainty.
Economic incentives can reinforce this tendency. When searches become inexpensive and immediate, agencies can conduct inquiries that once required more time and personnel. The lower cost of obtaining information makes additional searches easier to justify. Yet the benefits and risks are not necessarily distributed equally. The institution gains speed and investigative capacity, while the individual may bear the consequences of an erroneous match, an unjustified stop, or the exposure of personal movements.
This asymmetry also affects how institutional performance is evaluated. The number of searches completed, the speed of identification, and the volume of records processed are relatively easy to quantify. By contrast, the number of unjustified interventions prevented, the errors left uncorrected, and the consequences of inappropriate access are more difficult to measure. If administrative performance is assessed primarily by the former indicators, an organization may become more efficient at processing information without becoming correspondingly better at establishing facts.
The scale of access to surveillance data makes these questions particularly important. McMaken reports that ten Flock cameras in a Utah county were queried more than five million times between 2022 and July 2026, with nearly 97 percent of searches originating from law-enforcement agencies outside Utah. These figures do not establish that the searches were improper. They demonstrate how widely locally-collected information can circulate beyond its original setting. As access expands, meaningful oversight must address not only who can retrieve information but also the purposes for which it is used and the consequences of decisions based on it.
The issue, then, is not whether surveillance can produce useful information; it can. Nor does every error indicate deliberate misconduct or prove that a technology should be abandoned. The more precise question is whether institutions maintain a reliable distinction between information that warrants further inquiry and evidence sufficient to justify coercive action. Without that distinction, procedural activity can obscure uncertainty rather than correct it.
A meaningful response must therefore address the decision-making process, not merely the system’s technical performance. Access should be limited to legitimate purposes, and searches should leave auditable records. Automated matches should be independently verified before being used to justify consequential interventions. Decision-makers must consider contrary evidence rather than treating an initial alert as presumptuously authoritative. Individuals affected by inaccurate information should have effective means to challenge decisions and correct records. These safeguards matter because they interrupt the process by which an uncertain indication can gain lasting institutional authority.
The broader lesson extends beyond surveillance. Whenever an organization uses information to classify people and authorize action, it risks mistaking the outputs of its own procedures for independent knowledge of the reality those procedures are meant to describe. More information can improve decisions, but only if the institution remains capable of questioning its interpretations, acknowledging uncertainty, and correcting errors.
Digital surveillance promises security by enabling authorities to see more, but seeing more is not the same as understanding more, and an investigation is not proof that the suspicion prompting it was justified. The decisive question is whether institutions can distinguish a useful lead from established evidence and remain willing to revise their conclusions when the facts contradict them. Security depends not only on the capacity to collect information but also on the capacity to correct decisions made in its name.
