New tools deployed and offered to law enforcement agencies by Flock Safety, the largest US aggregator of automated license plate reader (ALPR) data from both government and private cameras, are moving Flock from data mining into profiling and pre-crime predictive policing. This marks the expansion to road travel of the profiling and predictive policing that was developed and has until now been applied primarily to air travel.

Flock’s data warehouse  includes billions of monthly records, each of which links a unique vehicle identifier (license plate number) to a precise date, time, and location. Flock boasts that it is now using artificial intelligence (i.e., more complex algorithms) to identify patterns and “surface” evidence of suspicious activity.

This moves Flock from a surveillance company and provider of investigative tools to a provider of suspicion-generating and predictive tools for “pre-crime” policing. This makes  a quantitative difference in degree of intrusiveness and danger, of course, but there’s also a qualitative difference between an investigation based on a lawful pre-existing basis for suspicion, and dragnet surveillance intended to generate a basis for new suspicion (that can in turn be used as the basis for further surveillance, search, seizure, detention, etc.).

Thanks to Jay Stanley for calling attention to these new Flock tools in articles on the ACLU website and and in his own Free Future newsletter.

There’s a  precedent for dragnet suspicionless collection of timestamped, individually identified movement logs and their use as inputs to an algorithmic suspicion-generating and predictive profiling engine. That precedent is the government’s compelled identification of air travelers, reporting of airline reservation data to the government, and use of this data to generate a predictive “threat score” for each air traveler.

The result of this completely misguided and inherently flawed exercise in pre-crime policing of air travelers has been the creation of the US government’s “bigotry in, bigotry out” no-fly and “selectee” lists of more than million names, mostly of innocent Muslims, used as the basis for extra-judicial denial of the right to travel by common carrier and more intrusive warrantless searches of air travelers selected by these lists and algorithms.

Let’s be clear: Flock doesn’t have any more pre-cogs than do the agencies trying to predict threats to aviation from airline reservation data. “Artificial intelligence” provides less probable cause than simpler algorithms for lawful suspicion or action, because with AI it’s impossible to say which specific inputs led to to the score or output, “this is suspicious”.

Courts should reject any claim that outsourcing surveillance or data analysis to Flock lets government agencies off the hook for using this data or the results of this profiling, or that a conclusory but perhaps hallucinatory output from Flock’s travel-profiling AI provides probable cause for lawful suspicion, search, seizure, or issuance of a warrant.