Product2 distinct publishers3 min readUpdated
Flock now asks officers to type a criminal case number before searching its plate-reader network. It told MIT Technology Review it does not verify the numbers.
The Product Desk · Product desk

Compiled by The Product DeskSomething wrong?How this is made
Flock, the police-technology company whose network includes some 120,000 automatic license plate readers around the US, announced changes last Thursday meant to stop officers from using the platform for illegal or illegitimate purposes [1][2]. The headline control is a requirement that anyone running a search enter a criminal case number, alongside software that flags abnormal searches [3]. Flock confirmed to MIT Technology Review that it does not verify those case numbers, so an officer can simply enter a bogus one [4].
That is the entire lifecycle of an attestation-based safeguard in three sentences. The control asks the user to assert a fact, records the assertion, and never tests it. For anyone acting in good faith it adds a field to a form. For anyone acting in bad faith it adds a field to a form.
The population this control is aimed at is not hypothetical. The Washington Post recently identified 50 cases in which officers misused systems from Flock and its competitors, often to stalk and harass women [5]. One woman in Wisconsin alleged that her officer ex-boyfriend searched for her car 179 times [6]. Another was being stalked by the chief of police, with nobody to report him to [7]. MIT Technology Review notes that officers have already lied to get around other Flock safeguards [8]. A safeguard whose only failure mode is "the user types something untrue" is being deployed against people who have demonstrated they will type something untrue.
The alternative is not exotic, and MIT Technology Review spells it out: require case numbers that match the police department's own records [9]. That is a more intrusive integration, and it is the reason it does not exist. Verification means writing connectors to hundreds of records systems, handling the department that has no queryable case management at all, and accepting that some legitimate searches will now fail. Attestation costs one text input and ships this quarter. The difference between the two is not diligence, it is a build decision, and the audit trail is where it shows up: an unverified number tells an investigator nothing later, while a matched one points at a real file [9].
The same pattern repeats in Flock's retention story. Flock says 90% of searches happen within a week of an incident, and the company recently changed its recommended retention period to seven days [10][11]. Recommended. Agencies can hold data for as long as they like, and because Flock mostly operates as a national network, police in one city or state can search data collected in another and retain it for months or years [11][12]. A recommendation is an attestation with the officer removed: a stated norm with no enforcement surface [13].
There is a version of this system with a real constraint rather than a stated one. MIT Technology Review's example is scoping the broadest access to emergencies, so that searches tied to an active Amber Alert can reach data from surrounding cities [14], with kidnappings and missing persons being the use case Flock cites more than any other [15].
Watch whether Flock ships records-system integration or leaves the field unverified, and watch the venue. Cities are already cancelling contracts and states are trying to pass laws limiting or banning plate readers [16]. Chad Marlow, a senior policy counsel at the ACLU, told MIT Technology Review that the best way to set limits on surveillance is not new Flock guidelines but new laws [17].
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Flock is a police-tech company known for its network of some 120,000 automatic license plate readers around the US.
Flock announced changes to its platform last Thursday, meant to prevent officers from using the platform for illegal or illegitimate purposes.
Flock's new practices include using software to flag abnormal searches and requiring searchers to enter a criminal case number.
Flock confirmed to MIT Technology Review that it does not verify the case numbers officers enter, so an officer can simply enter a bogus one.
One woman in Wisconsin alleged that her officer ex-boyfriend searched for her car 179 times.
Another woman was being stalked by the chief of police, with nobody to report him to.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Vendor-confirmed, two independent readings
The central factual claim rests on Flock's own confirmation to MIT Technology Review that it does not verify case numbers, and the surrounding change set is independently itemised by Techdirt from Flock's announcement, including quoted company language. Prior misuse is grounded in Washington Post reporting and named cases. Weaknesses: the two sources disagree on the misuse case count, the Techdirt body is truncated, and no source quantifies whether the network actually solves or prevents crime.
Network broadly deployed, safeguards partially adopted
Deployment of the underlying network is large and measurable (about 120,000 readers, roughly 20 billion plate images added monthly, access for any contracted agency). Adoption of the accountability controls is much thinner: abnormal-activity auditing was voluntarily taken up by only about a third of customers and becomes mandatory only later this year, the seven-day retention limit is a default rather than a cap, and offense filtering is opt-in per agency. Counter-adoption is also measurable in cancelled municipal contracts.
Reform framing outruns enforcement
The announcement is presented as preventing illegitimate searches, but the mechanisms are weaker than the framing: the case-number requirement is unvalidated by the company's own admission, the seven-day retention period is a recommendation and default rather than a limit with 30 days still standard, and the strongest control (abnormal-search auditing) had only one-third voluntary uptake. Both publishers independently identify this gap, so it is the vendor's positioning that is overstated rather than the reporting. The gap is not maximal because several changes are concrete and dated: mandatory MFA, third-party security review, offense filtering, and audit reviews already linked to officer arrests.
Reactive vendor reform under commercial and political pressure
Incentives are unusually legible on every side. MIT Technology Review reports that Flock's business model and roughly $8 billion valuation depend on leveraging cameras into a broad national network, so narrowing scope would threaten its pitch to police departments; Techdirt documents that the changes followed six weeks of negative press, senatorial scrutiny, and municipal cancellations, with the CEO apologising on television. The ACLU's stated position that no acceptable contract exists, and the CEO's reply that he will always differ from the ACLU, mark the advocacy incentives on the other side. Both publishers write in a critical register: one as an opinion newsletter, one as commentary.
High on the core mechanism, moderate on scale of harm
Confidence is high that the case-number requirement is unverified and that retention remains agency-controlled, because those come from the vendor and are corroborated across both sources. Confidence is moderate on the magnitude of documented misuse (50 versus 46 cases, differing scope) and on how much the announced controls will change behaviour, since mandates fall due later and no source measures crime-solving efficacy or post-change search volumes.
build
Flock cut retention to seven days and published the number that made 30 indefensible1 distinct publisher
product
Flock built the thing it said it could not do, then left the prompts on a public server4 distinct publishers
build
Flock's OS Investigate starts the search before there is a suspect1 distinct publisher
leadership
Longmont dropped Flock and kept the cameras: vendor churn is not surveillance policy1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 17, 2026
1 article · August 17, 2026