Flock Safety built an AI tool that tracks people, WIRED finds

Flock Safety has spent years telling the public its license-plate cameras "cannot recognize, identify, or track individuals." Code uncovered by WIRED behind the company's law enforcement login pages tells a different story.

The reporting reconstructs a system called OS Investigate, internally known during development as Nightshift, that is built specifically to identify drivers and follow vehicles based on their movement patterns over time. That capability sits squarely in the territory Flock has repeatedly said its technology avoids.

OS Investigate arrives with 69 prewritten prompts officers can select, tweak, or run as-is, functioning as a guided interface for querying police data with natural language. Behind those prompts, the system can call on 45 separate tools spanning plate scans, camera metadata, arrest records, case files, dispatch logs, ballistics results, and commercial databases containing Social Security numbers and other personal records.

From plate to person

The practical effect is a pipeline that converts a license plate number into a name, a home address, and a list of relatives, pulling from 911 dispatch logs, case files, and third-party identity brokers along the way. Flock's camera network already logs vehicle movement across more than 6,000 communities in the United States, giving the new AI layer a large dataset to query.

Flock told WIRED that OS Investigate remains in testing with a limited set of law enforcement partners and has not been broadly deployed. The company did not dispute that the system was found on its own infrastructure, and WIRED says it reconstructed the tool by examining code served through Flock's public-facing login pages rather than through a leak or breach.

Flock's own marketing has long insisted its cameras "cannot recognize, identify, or track individuals."

The disclosure lands as Flock already contends with bipartisan criticism, documented cases of officers misusing its camera platform, and vandalism against its hardware in several cities. Separately, the company announced an "Audit Assistance" feature on August 13, 2026 designed to flag abnormal searches, which Flock has explicitly described as not built on machine learning or AI, a distinct project from OS Investigate rather than a response to it.

For a company whose business model depends on convincing cities, police departments, and residents that its surveillance network respects individual privacy, a tool purpose-built to unmask and follow specific people is a hard claim to reconcile with past assurances. Flock has not said when, or whether, OS Investigate will move beyond its current pilot phase.