Flock's police search tool can now hunt a specific person across an entire camera network based on nothing but a written description.
Flock hadn't detailed the feature publicly, so WIRED reconstructed it directly from the code the company ships to a police officer's browser. According to WIRED's rebuild, the interface lets an officer type a description of a person and have the AI scan footage from multiple cameras at once for a match. The company did not make this interface available for outside review before WIRED's investigation. That WIRED had to pull the tool apart from client-side code, rather than being shown it, is itself part of the story.
This is a meaningful jump for Flock, whose name has mostly been tied to license-plate reading. Matching plates is a narrow, mechanical task. Matching a person to a text description across camera feeds is a much fuzzier, more consequential kind of AI judgment call, and it puts Flock in the same territory as facial-recognition vendors that have drawn years of scrutiny over accuracy and bias.
The pattern is familiar: the public learns what a policing AI tool actually does not from a spec sheet, but from someone taking apart the code after the fact.