Exposed Code Reveals Flock Safety AI Tracking Drivers Without License Plates

In brief

• Flock Safety‘s OS Investigate includes 69 prewritten prompts that connect police AI to vehicle scans, criminal histories, and commercial data holding Social Security numbers and family contacts.
• According to Wired, 41 prompts require no administrative approval, and 14 operate without a license plate, name, or vehicle description—relying solely on time, location, and movement patterns.
• An associate-finding prompt flags secondary vehicles appearing near a target car within two minutes, running counter to Flock’s public claims that it does not monitor individuals.

Code left accessible on login pages for Flock Safety has revealed an artificial intelligence surveillance system capable of tracking motorists based strictly on their driving patterns across more than 6,000 U.S. communities.

Tech publication Wired detailed the software’s capabilities after obtaining source code embedded directly within web files that Flock served to visitors on its public login portal.

According to the firm, the application—originally developed under the name Nightshift and now designated OS Investigate—remains under active development while undergoing trial runs with a limited selection of police agency partners.

The program operates through a chat-style interface where officers select from a preset catalog of 69 prompts or enter customized text commands. One integrated search function compiles potential “witnesses” by identifying automobiles recorded most frequently within a specific neighborhood during the preceding two weeks.

A secondary prompt identifies individuals with more than two arrests within a two-year span—excluding drug-related charges—and automatically assembles profiles for the top three subjects. Additionally, a “workup” function converts a target’s name and birth date into comprehensive profiles detailing family members, telephone numbers, and web accounts harvested from law enforcement files and private data aggregators.

Exactly 14 pre-formulated prompts function without requiring a license plate number, individual name, or physical description. Instead, an investigator designates a geographic location, timeframe, and behavioral pattern—such as vehicles visiting three retail establishments over three days or multiple financial institutions in a single week—and the platform outputs matching individuals. While an automated filter removes commercial buses and delivery trucks, standard civilian drivers making identical stops remain flagged in the results.

The underlying code explicitly outlines the mechanism for establishing suspect associations. The system measures how frequently secondary vehicle plates are logged by identical cameras within a two-minute timeframe of a primary target vehicle, highlights vehicles appearing three or more times with a confidence rating exceeding 0.75, and outputs up to 20 identified individuals.

Automated License Plate Readers traditionally capture optical character recognition data—such as plate numbers, timestamps, and camera coordinates—to check against police watchlists. By combining real-time location logs with commercial data brokers that pull together public records, phone directories, and Social Security numbers, AI-powered tools allow investigators to link physical vehicle movements with personal identity profiles.

“I don’t know how else to say this, but this sounds completely insane,” former Pawtucket, Rhode Island, police officer Noel Pichardo told Wired after examining the query list. “I don’t know how anyone can argue against the idea that Flock literally tracks people.”

Jay Stanley, a senior policy analyst with the ACLU, observed that the narrow margin between the full technical potential of Flock’s hardware and its current operational deployment leaves “very little space… like China.”

Flock has consistently asserted in public statements that its platform “cannot recognize, identify, or track individuals” and is “not for watching people.” Company spokesperson Paris Lewbel characterized OS Investigate as a standalone software tool distinct from its license-plate reader hardware, designed to streamline data processing across records already maintained by law enforcement agencies. Lewbel added that the tool’s features could be modified substantially prior to broader deployment, though Flock did not challenge the accuracy of Wired’s findings.

Public opposition has manifested in direct physical actions. During June and July, anti-surveillance activists targeted Flock hardware nationwide through spray-painting, taping over lenses, and physical destruction, with The Guardian documenting at least 33 individual incidents spanning 23 states. This backlash aligns with documented instances of internal misuse; research released by the Institute for Justice highlighted over two dozen occurrences where police personnel resigned or faced criminal charges for allegedly employing license-plate readers to stalk current or former romantic partners.

Political resistance has extended to Capitol Hill, where Rep. Thomas Massie is preparing legislation designed to strip federal grant funding from law enforcement departments that implement the surveillance network.

While privacy researchers have developed AI-generated patterns to blind plate-reading software, major tech firms continue pushing broader surveillance tools, with Meta securing patents for cameras capable of recognizing faces and recording individual behavior.