This Is Flock’s AI Search Tool for Cops

Details

Flock Safety, whose license-plate readers operate across thousands of US cities, has expanded into AI-powered person searches: officers type a natural language description and set cameras across a defined area to scan continuously for matching individuals. WIRED reverse-engineered the JavaScript Flock sends to officers' browsers to produce a detailed account of how the system actually works. The investigation follows a year of mounting controversy. A Texas deputy used Flock to run more than 83,000 camera searches for a woman who had sought an abortion. Illinois found Flock had given federal immigration agents access to state camera data in violation of state law. Flock responded in August 2026 with reforms: shorter default data retention, mandatory case codes per search, and automated audit logging. WIRED's code analysis found a critical gap: the system flags and records searches containing sensitive content — race, religion, political/cultural expression — but only blocks the first two categories outright. Searches touching political or cultural expression get a speed bump, not a stop, which experts note is precisely the category with the strongest First Amendment protections. Flock's own documentation acknowledges results may be inaccurate and shifts liability to the officer, not the company. Experts in automated screening say no AI-based person-identification system operating at this scale is reliable, and Flock's false-positive rates cannot be independently verified by anyone outside the company.