Kern County Sheriff | License plate readers

Two policies, four years apart

The Kern County Sheriff’s Office publishes its policy manual by section, which is more than most California agencies do, and Section J holds two documents about licence plate readers.

The older is Policy J-3800, effective 3 March 2020. It governs two commercial products: Thomson Reuters’ CLEAR, and Vigilant Solutions’ LEARN, which the policy describes as providing “the ability to query ALPR data obtained from commercial and other law enforcement sources”. Its definitions section ends a paragraph with a sentence set in bold:

The Kern County Sheriff’s Office does not deploy ALPRs.

In 2020 this department read plate data. It just did not collect any. What it bought was access to databases assembled by other people, some of them commercial: repossession contractors, insurance investigators, other agencies. That is a real and often overlooked way a sheriff’s office gets into plate surveillance, and it leaves almost no local trace, because there is no camera on a pole for anyone to notice.

The newer is Policy J-3810, effective 24 July 2024, and it governs the department’s own Flock cameras.

Both were still published, side by side on the same index page, when we read them in August 2026. J-3800 carries no revision after 2020.

This is not a contradiction and we are not going to write it as one. A policy describing 2020 practice is not falsified by a 2024 deployment. What it is, is a dated record of a department crossing a line that a lot of agencies have crossed in the same direction: from querying other people’s plate data, to running its own network.

What the 2024 policy permits

Two sentences in J-3810 are worth reading against each other. The first is in the general use section:

Flock may be used in conjunction with any routine patrol operation or criminal investigation. Reasonable suspicion or probable cause is not required before using Flock ALPR.

That is the standard for running a search, and it is written down plainly, which is better than leaving it to inference. It is also, stated plainly, no standard at all: a deputy on ordinary patrol may query the system without needing a reason a court would recognise. The check the policy relies on instead is after the fact, and it is real but limited: the administrator “shall conduct periodic audits of queries”, and misuse can bring “Criminal and Civil Liability” and disciplinary action “up to termination”.

The policy also describes what the system can do beyond reading a plate. Vehicle Description Alerts, built on the vendor’s vehicle-fingerprint technology, send notifications when a vehicle matching a set of filters passes a camera, and are “most used when a license plate is not known”. A Look Up Tool lets agencies opt their cameras into a searchable network, and “An agency can contribute on a statewide level”. Custom hotlists let a user create alerts for specific vehicles that would not otherwise alert, with entries expiring after no more than thirty days.

And on what the public may see of any of it:

The data viewed and utilized may contain confidential information, therefore it is not open to public review.

The sentence worth crediting

The second sentence to read carefully is in the section on releasing data:

Flock ALPR data is prohibited from being shared with out-of-state agencies and federal agencies.

We should say plainly that this is better than what we have found elsewhere. California law has restricted this sharing since 2016, and the California State Auditor nonetheless found, in Report 2019-118, that one department three counties north had authorised sharing its plate images with 982 outside entities. This department wrote the prohibition into its own policy, added a procedure for California requests that requires a written purpose, a commander’s approval and a retained copy of the request, and published the whole thing on its website where anyone can read it.

What we have verified is the policy. We have not verified the practice, and nothing on this page should be read as saying we have. A written rule and a followed rule are different objects, and the only way to tell them apart is an audit or a records request. The audits J-3810 requires would be a good place to start, which is why they are in the request on this site.

What is missing

Neither policy says how long a plate read is kept. The thirty-day cap applies to custom hotlist entries, which is a different thing entirely: it limits how long an alert stays armed, not how long the record of your car passing a camera survives.

Neither policy says how many cameras there are, or where. And nothing we could reach says who approved them. Kern County’s own website answers automated requests with a bot wall, and the Board of Supervisors publishes its archive through a portal that turns us away, so the decision that created this network is not something we have been able to look at. That is a gap in our record rather than a finding about the county, and it is first on the list of things a person with a browser could fix.