Police AI Tool Searches Cameras for People Based on Text Descriptions

Flock Safety's new AI search tool lets officers define an area and have cameras continuously look for individuals matching a written description. The company has added data retention limits and auditing after criticism, but experts say the system's accuracy is unproven and may not prevent misuse. WIRED analyzed the software and created a mockup of the interface.
The tool operates through a browser-based interface that WIRED reconstructed from code sent to officers. Searches can run continuously across multiple cameras within a defined geographic area. Flock's August policy changes mandate case codes for every query and automated auditing, though the company's servers handle much of the search logic, limiting external oversight.
The system evaluates written descriptions against content categories including race, religion, and political expression. Only political, social, and cultural expression triggers a warning rather than a block. Experts note this category enjoys the strongest First Amendment protections, yet Flock treats it as the least restrictive. The company acknowledges results may be inaccurate but places responsibility on officers.
This technology could reshape how police conduct surveillance, potentially enabling broad, automated monitoring of individuals based on subjective descriptions. Civil liberties concerns may intensify if accuracy remains unverified