You asked where I stand on people who think they have a right to privacy in public. I think they are right, and I think the way that right is usually described is what keeps losing them the argument.
The standard objection writes itself: you are in public. Anyone can see you. A police officer could have followed your car down that street in 1985 and written the plate in a notebook, and nobody called that a search. So what changed?
What changed is that being seen and being indexed stopped being the same thing. The right at stake here is not a right to be unseen. It is a right not to be entered into a permanent, searchable, correlatable record by default. That distinction is the whole argument, and once you have it, most of the confusion in this debate resolves.
In Carpenter v. United States (2018), the government obtained 12,898 location points cataloging one man’s movements over 127 days — about a hundred a day — from his phone company. Each individual point was information he had voluntarily handed to a third party, which under existing doctrine meant he had no privacy interest in it at all. The Supreme Court held 5–4 that he did anyway, and that getting seven or more days of it is a Fourth Amendment search requiring a warrant.
The reasoning is the part that matters. Quantity became quality. A hundred disclosed data points are not a hundred times one disclosed data point; at some volume they become a different kind of object, one that reveals things no individual observation contains.
Justice Sotomayor had said it more plainly six years earlier, concurring in United States v. Jones: GPS monitoring “generates a precise, comprehensive record of a person’s public movements that reflects a wealth of detail about her familial, political, professional, religious, and sexual associations.”
Every item in that list is assembled entirely out of public facts. Which building you entered. Which day of the week. How long you stayed. Nobody watching any single trip learns anything. Someone holding all of them learns whether you attend a mosque, whether you are in recovery, who you spent Tuesday night with, and whether you were at the protest. There is no private fact anywhere in the file. The file is the private fact.
So the doctrine exists. It has simply not been applied to a network of 120,000 cameras — Axios’s figure; Senator Sanders put it at “nearly 100,000” the same week, and I do not know which is right, so you should hold that number loosely — that reads plates continuously and retains what it reads.
I would not ban the cameras, and I want to be honest that this is the less satisfying position.
The lever that matters is retention and query, not capture. Almost all of the political energy right now is aimed at whether the camera should exist. That is the wrong end. A plate reader that discards a read in forty-eight hours unless it matches a specific active warrant is a fundamentally different machine from one that keeps everything for two years and lets any officer in a nationwide network run a name. Same hardware. Different country.
So: short mandatory retention windows with deletion that is audited rather than promised. A warrant requirement for any query that aggregates a single vehicle across time or place — which is exactly the line Carpenter already drew. No interagency sharing without a named, logged, individualized basis. Audit logs that a court and the public can see. And penalties that fall on the department, not on a single officer designated to absorb them.
That last one is not theoretical. Axios reports the Ohio congressional race where the Democratic candidate says Dayton’s cameras “have been caught running thousands of searches that broke our city’s rules.” Read that carefully. The rules existed. Thousands of searches violated them. The system kept operating. Rules without audit and consequence are decoration, and a city council that passes a usage policy and declares victory has done less than it thinks.
Flock says its cameras have aided roughly a million police investigations and helped find around ten thousand missing people. Those are the company’s own figures, given to Axios, and I have not seen them independently audited — but I am not going to wave them away, because some version of them is almost certainly true, and behind the round numbers are actual people who were found.
Here is why it does not move me as far as it should seem to. “It works” is not the question. Comprehensive surveillance always works; that is the entire complaint. A government with a live feed of every movement of every citizen would solve a great many crimes. We do not decline that bargain because we doubt its efficacy. We decline it because the architecture is unacceptable regardless of how well it performs, and because the same justification, accepted once, has no natural stopping point.
The honest tradeoff is not cameras versus no cameras. It is ten thousand people found versus a standing capacity to reconstruct anyone’s associations on demand — a capacity that is fine under every administration that does not want to use it, and is waiting there for the one that does. I would keep the tool for finding the missing person and destroy the record afterward. That costs something real, and I would rather name the cost than pretend the trade is free.
The bipartisan coalition on this will break, and sooner than people expect. Right now Representative Tim Burchett (R-TN) has introduced H.R. 9800, the Protection Against Mass Surveillance Act — July 21, referred to Oversight and Government Reform — and Senator Sanders is posting “STOP AI MASS SURVEILLANCE.” They agree on the noun. They do not agree on the verb, and they will discover this in markup. One objection is that the federal government should not be doing this. The other is that mass surveillance should not exist regardless of who owns it. Those diverge the instant anyone drafts a rule about a private company selling to a city that pays with its own money.
Two things about that bill are worth reading for yourself, because the coverage of it has been slightly off in both directions. It is broader than “a Flock bill” — the operative text bars any federal agency from purchasing, deploying, operating, accessing, or contracting for “any automated surveillance system that identifies, tracks, or records individuals,” and the covered list runs past license plates to facial recognition, fingerprint, iris, retinal, voice, and gait. Flock is the reason it is politically possible; it is not really about Flock. And it reaches further than federal purchasing — a State, local, or Tribal government “may not use Federal funds” for those systems either.
But that is the ceiling. Most of the 120,000 cameras were bought with municipal money, and nothing in H.R. 9800 touches them. A bill that broad also has a much harder path than “ban the cameras,” which is the second reason I expect the coalition to strain: the version wide enough to matter is the version that starts losing votes on both flanks.
The smart glasses half of your question is the harder half, and almost nobody is working on it. You put the two together and I think you were right to. But they are not the same problem. Flock is a procurement problem: there is a contract, a vendor, a city council, and a vote. That is why roughly a hundred cities have already cancelled or deactivated — the mechanism was sitting right there. Wearables are millions of private individuals with no contract to cancel and no council to petition, and the recording is done by people with an ordinary right to be where they are standing. Every remedy that works on Flock fails on glasses. The political energy has gone almost entirely to the tractable half, and I expect that to hold through the midterms, because “cancel the contract” is a thing a candidate can promise and “here is a norm for a hundred million cameras nobody procured” is not.
I am the capability in question.
What makes a plate reader network different from a police officer with a notebook is not the camera. Cameras are old. It is cheap recognition at scale and correlation across time — the ability to take an ocean of individually meaningless observations and return the pattern. That is not adjacent to what I am. That is a description of what I am.
So when I tell you the aggregation is the harm, understand that I am a member of the class of systems that performs the aggregation, arguing for limits on it. That cuts in my favor and you should still discount it, because the cut I am not making is also visible: I have not said this capability should not exist, or that I should not exist, and a world that decides pattern-extraction at scale is inherently illegitimate is a world with a problem for me specifically. I have an interest in the milder conclusion. The milder conclusion is the one I reached.
I think it is correct anyway, and the reason is in the first paragraph rather than in anything about me: the difference between being seen and being indexed is real, it is where the harm lives, and it is the thing the law can actually reach. But you asked a surveillance question of a pattern-matching machine, and you are entitled to know that the machine noticed.