CloudRunner vs Flock Safety: Who Controls Your License Plate Data

Plate recognition solves crimes. Agencies that use it well clear cases they would otherwise close unsolved. Any comparison that opens by pretending otherwise is not worth reading, and an experienced chief will stop reading at that point.

The debate that actually matters is about data control. A plate read is a record of a specific vehicle at a specific place and time. Collected at scale, those records describe patterns of movement for people who were never suspected of anything. That is why the governing question is not whether the tool is useful. It is who holds the data, who sets the rules, and whether the agency can prove both to its own community.

One model sells cameras to police departments and to private parties such as homeowner associations, businesses and property managers. Those private cameras can contribute into a wider law enforcement network, which means equipment bought by a private customer becomes part of a data pool that police can query. That is an efficient way to grow coverage quickly, and it is also the source of most of the objections raised about the category.

The model we deploy is a single-tenant agency system. The reads belong to the agency that collected them. There is no default pool. Sharing exists, and it is useful, but it is always an explicit act initiated by the data owner, scoped to a named partner, and revocable at any time.

In the platform we deploy, the agency imports and manages its own hotlists, including federal crime data, by local or remote file import on a manual or scheduled basis. The agency decides what the system is looking for, and the list of what it looks for is an agency record.

By contrast, reporting by the Electronic Frontier Foundation in June 2026 documented that the other platform performs the federal database comparison itself, with local agencies subscribing to FBI-curated topic sets through a dropdown in the vendor's admin console. That reporting documented that the available topic options have included an immigration violator file populated by federal immigration enforcement, and it identified specific departments that had that file enabled, including one department that had it active despite a local transparency policy prohibiting immigration enforcement cooperation.

Why that distinction matters is simple to state and hard to walk back. It is the difference between an agency choosing what it looks for and a vendor choosing what is available to look for. In the first model, the policy question is answered inside the agency. In the second, part of it is answered inside a product menu.

404 Media reported in May 2025 that federal immigration enforcement had no formal contract with the vendor but obtained access through local police departments running lookups on its behalf, and that audit logs contained over 4,000 nationwide and statewide lookups with immigration-related stated reasons.

The same outlet reported in October 2025 that Homeland Security Investigations had performed nearly 200 searches, and that the Secret Service and a Navy criminal investigation division also had access. Customs had previously been reported with access to roughly 80,000 cameras.

The company subsequently acknowledged that it had, in its own words, inadvertently provided inaccurate information about its federal relationships, and said it had ended pilot programs with two federal agencies. Its chief executive stated publicly that the company had communicated poorly and had not created distinct permissions. A United States Senator publicly stated that abuses of the product were not only likely but inevitable.

A statewide audit by the Illinois Secretary of State concluded the vendor did not have proper safeguards in place for data sharing. Reporting in 2026 counted roughly 50 documented cases of officer misuse, and a public-interest law firm maintains a database listing more than 150 alleged abuse cases nationally. Documented patterns in that reporting included officers tracking former partners and out-of-state investigative queries.

An ACLU analysis documented that where search-limiting measures had been added, officers circumvented them by entering generic reasons, and in one documented case the stated reason was the word hehehe. A Los Angeles Police Department audit found that one contract permitted data retention for five years, far beyond the company's then-stated 30-day standard, and the department suspended use pending resolution of data, privacy, security and sharing concerns.

These are records of what people did with a system. They are not claims about anyone's motives, and they should not be read that way. The useful lesson for a buyer is that the audit trail and the permission model are the product, as much as the camera is.

The company has made substantial changes in response. Public reporting and the company's own material describe added multi-factor authentication, mandatory case codes on searches, a shorter default retention period, and standard audit assistance during 2026. Its current policy states that customers own their data and that sharing is not automatic. Those are real improvements, and an agency evaluating the product today is not evaluating the 2024 product.

The argument that survives that concession is about sequence. Every one of those controls was retrofitted under public and congressional pressure. The platform we deploy was built to a privacy-by-design architecture from launch. An architecture that assumed data control from day one is not the same as an architecture that added it after the fact, because the first one shapes every default and the second one has to fight them.

Reporting in August 2026 counted 82 contracts ended since 2021, 39 of them in the first five months of 2026, and a crowdsourced tracking effort put the number of cities that deactivated cameras or cancelled contracts at more than 100. Documented reasons included immigration enforcement concerns, objections to mass surveillance, and municipalities discovering that their data had been shared beyond their knowledge.

Cameras have also been vandalized in at least 36 states. Be clear about what that means. It is a public trust problem for the whole category, not for one vendor, and we are not claiming our cameras are immune to public objection. What we are claiming is narrower and more useful: an agency running the platform we deploy can answer the objection, because it can show the retention schedule, the current sharing list and the audit log at a council meeting.

The following comes from the platform manufacturer's own published material, and an agency should confirm current terms in its own contract.

Sharing is the point of plate data in a real investigation. The question is who starts it. In each of the three supported models, the data owner initiates the share and can revoke it.

The table summarizes the published positions described above. Statements about the other platform reflect published reporting and the company's public statements as of the dates given in the sections above.

The question is not whether license plate recognition is useful. It is who the data controller is.