ALPR learning experiment
Watch a fictional vehicle leave a handful of real-looking records behind. Then decide what those records actually tell you.
Fictional vehicle · fictional locations · no real person is being tracked · about 45 seconds plus one question
What do those records actually tell you?
All of those stories could fit the same data.
The data can reveal a routine without revealing its meaning.
In a documented 2026 Plymouth, Minnesota case, an incorrect stolen-plate database entry contributed to Flock alerts and a police stop of an innocent driver. Flock told CBS Minnesota that an alert should be one part of an investigation, not the entire basis for a stop.
The camera saw a vehicle.
The network found a routine.
The search found a match.
None of that proved the story was true.
Observation ≠ inference ≠ proof.
Flock's current LPR policy says LPR data can include plate and vehicle images, plate number and state, vehicle characteristics such as color and make, date, time, and camera location. It says queries are audit-logged and LPR data is hard-deleted on a rolling seven-day basis by default, subject to customer law or policy. Read Flock's policy ↗
The ACLU argues that networked ALPR systems can create location-surveillance risks because repeated observations and data sharing can expose patterns of movement at scale. Read the ACLU analysis ↗
The Plymouth sequence shown in the experience is based on a 2026 CBS Minnesota report involving an incorrect stolen-plate database entry and subsequent Flock alerts. Read the report ↗
The vehicle, locations, timing, movement pattern, and interface are fictional. They illustrate the distinction between an observation, an inference drawn from observations, and proof.