Flock Safety has built an AI tool that identifies drivers and tracks vehicles by their movement patterns alone, drawing on a camera network that logs traffic in more than 6,000 communities. The software, originally called Nightshift and now branded OS Investigate, ships with 69 prewritten prompts, 45 tools that reach into arrest records, 911 dispatch logs, ballistics results, and commercial identity databases containing Social Security numbers, dates of birth, and lists of relatives. Flock has told the public for years that its technology "cannot recognize, identify, or track individuals." The code tells a different story.
The prompts and interface were reconstructed by WIRED from more than 450 files served by Flock's own login pages to anyone who loaded them. The files were sufficient to rebuild portions of the authenticated police-facing application from scratch, including its search fields, tool calls, and controls. Flock did not dispute the capabilities described and characterized OS Investigate, through spokesperson Paris Lewbel, as "a separate product from Flock's license-plate reader technology, designed to help investigators work across information their agencies already have access to."
The original Flock transaction was narrow: a camera reads a plate, checks it against a wanted list, and discards the rest. OS Investigate inverts that arrangement. An officer no longer needs a plate, a name, or a crime to begin — they supply a place, a stretch of time, and a pattern of behavior, and the system returns the drivers who fit. Of the 69 prewritten prompts, 19 describe hunting for patterns rather than looking up a specific record, and 14 require no plate, name, or description at all.
Key facts
- 01Flock's OS Investigate (formerly Nightshift) ships with 69 prewritten prompts and 45 tools spanning arrest records, 911 dispatch logs, and commercial identity databases.
- 02The system draws on a camera network in more than 6,000 communities and can surface a driver's 'associates' by counting plates seen within a two-minute window at the same cameras.
- 0314 of the prewritten prompts require no plate, name, or description — just a location, a timeframe, and a behavior pattern.
- 04The associate-ranking tool defaults to a 0.75 confidence threshold and returns up to 20 vehicles from a single input plate.
- 05The default justification form requires an officer to type a reason but checks only that any accompanying case number contains at least three characters.
One canned prompt reads: "Find me witnesses based on vehicles most seen in [neighborhood] during [last 14 days] during [daily timeframe]." Another instructs the system to list everyone arrested more than twice in two years for "any offense," map where they live, pull calls for service at their homes, and "do a workup on the top 3 individuals." A "workup," in Flock's terminology, is a one-command background check that returns vehicles, prior suspect listings, relatives, phone numbers, and online accounts.
The behavioral searches lean on filters that make suspicion look like math. One group of prompts asks for vehicles that visited three or more retail locations in a city in three days, or multiple banks in a week, or multiple gas stations between midnight and 5 am. A built-in filter strips out buses, semi trucks, work vans, and trailers, so a delivery driver making the same rounds is excluded while an ordinary driver making similar stops remains on the list. Similar queries flag vehicles that left one city and returned within a week, made repeat roundtrips over 14 days, or passed through three areas in sequence.
The associate-ranking tool is more explicit. Given a single plate, the software counts how often other plates appear at the same cameras within a two-minute window of the target, and returns any vehicle that shows up next to it three or more times within a confidence threshold that defaults to 0.75. An officer supplies one vehicle. The system hands back up to 20.
Noel Pichardo, a former Pawtucket, Rhode Island police officer who was briefed on Flock during a departmental pilot and who reviewed the prompts, said the design contradicts Flock's public messaging. Pichardo testified against the technology before Rhode Island state lawmakers in 2024, weeks after serving a 30-day suspension for criticizing Flock and department leadership in a local newspaper. He resigned as the department moved to fire him following what he describes as retaliatory investigations.
Not every reviewer sees a problem. A detective at a California law enforcement agency, granted anonymity because he was not authorized to speak publicly, said his department would adopt the tool. He noted that Flock had already closed many vehicle-theft cases for the agency and that officers had worked with similar pattern-matching systems for years. He acknowledged the witness-finding prompts made him slightly uncomfortable.
The audit trail is thin. Before running a search, the software can ask an officer for a justification, and the code shows that by default a reason must be typed — but the form places no requirements on what that reason says or how long it must be. If the department requires a case number, the form checks only that the field contains at least three characters. Whether Flock applies additional server-side checks is not visible from the client code.
Flock says the product is still in development, tested with a small group of law enforcement partners, and that its capabilities may change before a broader release. The company also faces bipartisan political pressure, a growing record of officers caught misusing its platform, and vandalism campaigns that have left cameras sawed off and lenses painted over in cities across the country.
The commercial logic for Flock is straightforward. A plate-reader network monetizes on hardware and subscription; a reasoning layer that turns that same network into a pattern-of-life search engine monetizes on outcomes police care about — closed cases, identified suspects, mapped associates. It also raises the switching cost for any department that adopts it, because the value is no longer in the cameras but in the accumulated behavioral graph. That business shift is what makes OS Investigate the more important story than any single prompt inside it: the surveillance product category has moved from lookup to inference, and the constitutional and procedural guardrails written for the lookup era — warrants tied to specific vehicles, specific crimes, specific timeframes — have no natural purchase on a system that generates its own suspects from a map and a clock.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
