Controversial surveillance technology company Flock Safety has long insisted that its automated license plate readers (ALPRs) canât recognize or track peopleâs faces. But they may as well be able to, thanks to a new AI tool which can reportedly identify vehiclesâand their driversâbased on several other variables.

The new tool, called OS Investigate, is currently being tested with a small group of law enforcement agencies, according to a Wednesday report from Wired. Using code found on the Flock websiteâs login portal, Wired reporters and an independent security and privacy researcher who goes by the pseudonym Buchodi were able to reconstruct a mock-up of the experimental toolâs interface (without being able to access it directly).

The interface showed that OS Investigate lets officers search for vehicles via Flockâs network of cameras using natural language prompts. The system comes with sixty-nine prewritten prompts, Wired reports, each of which officers can edit before clicking submit. One such prompt tells the system to find âwitnesses based on vehicles most seen in [neighborhood] during [last 14 days] during [daily timeframe].â The investigation also found that while the system asks officers to provide a reason for making a particular search, it doesn’t appear to require typed reasons to be a certain length or include any particular details.

In contrast to the original business model for Flockâs ALPRs, which are supposed to check recorded license plates against lists of vehicles already suspected of involvement in a crime, the new system starts with a location, a timeframe, and/or a pattern of movement, and identifies vehicles from there. It also filters out buses, semi-trucks, business-owned vans, and trailers, according to Wiredâmeaning the searches are focused on privately owned and operated vehicles.

A Flock spokesperson described Investigate OS as âa separate product from Flockâs license-plate reader technology, designed to help investigators work across information their agencies already have access to,â according to a summary of the statement by Wired. The spokesperson also reportedly said that the toolâs capabilities and interface may change before it’s generally released to the public.Â

On its website, Flock writes that its ALPRs do ânot identify people by their facesâ or âidentify who is inside a vehicle.â By integrating with 911 dispatch logs, arrest records, and other public safety and commercial databases, however, the new tool reportedly makes it possible for officers to connect the dots between license plates and their drivers, as well as people in drivers’ social networks.

Flock has come under fire following a string of revelations of officers abusing its technology. Earlier this month, the Washington Post reported that at least fifty officers have been charged with or accused of misusing Flock cameras, in many of those cases to stalk women. The company said last week it would introduce new safeguards for its cameras in response to public backlashâwhich has manifested in widespread vandalism of its camerasâincluding a new tool that flags suspicious user behavior, which law enforcement customers will reportedly be required to adopt by the end of this year.

The controversy arrives at a time of rapid adoption of new AI tools among police departments across the U.S. Flock Safety and other self-described public safety tech companies, like Axon Enterprise, have been pushing new AI tools as a quick-fix solution to help law enforcement agencies accomplish more as some departments face staffing shortages. In lieu of comprehensive federal regulation, however, those agencies have, for the most part, been left to their own devices to figure out how to use the technology safely and ethically.