The Phantom Police Department Watching Real Streets

What happens when a fictional police department gets the keys to a city’s real surveillance network? For residents in several U.S. cities, this wasn't a hypothetical question—it was a corporate sales tactic.

Public records recently revealed that Flock Safety, a prominent automated license plate reader and surveillance company, created a phantom entity called "Flock City PD." This fake department wasn't used to solve crimes. Instead, it was a tool used by Flock employees to demonstrate the capabilities of their AI surveillance software to real law enforcement agencies. To make the pitch compelling, they didn't use simulated data; they tapped into live cameras in places like Dunwoody, Georgia, and Bryan, Texas, scanning real people going about their daily lives.

The core of the demonstration was a feature called "FreeForm" search. Unlike traditional systems that require a specific license plate number, FreeForm uses AI image recognition to allow semantic searches of video feeds. Audit logs showed that Flock employees queried the live feeds for highly specific and often sensitive subjects. Searches included "crowds," "person wearing a mask," "white truck with a trump sticker," and even "vehicle a pretty girl would drive."

When confronted, Flock defended the practice, stating that these invasive searches were actually designed to prove the system's ethical guardrails. According to the company, querying terms like "Star of David" or "large group with signs" was meant to show police officers how the software's moderation tools work. In some cases, the system successfully blocked the search (like the "pretty girl" prompt); in others, it issued a warning about First Amendment protections before allowing the search to proceed.

However, testing safety features on unwitting citizens sparked significant backlash. Residents argued that their privacy was being violated to serve as a live testing ground and sales prop for a private tech company. They were uncomfortable with the idea that a walk in the park or a drive to the grocery store could be commodified into a software demonstration.

Following public outcry, Flock has since isolated the "Flock City PD" demo environment, ensuring it no longer accesses real-world cameras or shares data with real law enforcement agencies.

This incident highlights a growing tension in the era of AI surveillance. As computer vision transforms physical spaces into searchable databases, the line between demonstrating technological capability and infringing on civil liberties becomes razor-thin. It forces us to ask: as AI tools become more powerful, who oversees the companies testing them in our backyards?

Key Points

• Flock Safety created a fictional 'Flock City PD' to demo its AI software to real law enforcement.
• Employees used the AI to perform semantic searches on live public cameras in real cities.
• Searches included sensitive political and personal descriptors to demonstrate the software's safety guardrails.
• Public backlash over privacy concerns forced the company to isolate its demo environment from live cameras.

Why It Matters

As AI image recognition makes physical spaces as searchable as the internet, the ethical boundaries of how these tools are tested and sold directly impact everyday civil liberties.

Sources:

• ‘Flock City PD:’ The Fake Flock-Owned ‘Police Department’ That Searched Real Cameras for Real People — 404 Media