Once Surveillance Cameras Are Installed, Cities Can't Remove Them: The Lock-In Dilemma of Flock Safety

The Event: Silent Expansion of America's Surveillance Infrastructure

Across streets, parking lots, and highways throughout the United States, an automatic license plate recognition (ALPR) system produced by startup Flock Safety is spreading at nearly invisible speed. According to reports, over 100,000 units have been installed nationwide, with the vast majority coming from Flock.

These seemingly innocuous surveillance cameras perform a simple task on the surface—recording vehicle license plates of passing cars. But their actual capabilities far exceed this description. Law enforcement officers can input natural language queries like "pickup truck with scratches on the left side carrying an ATV in the bed" or "green sedan with American flag stickers," and the system automatically matches against a massive image database, listing matching vehicles and their drivers.

More troubling is that Flock cameras aren't limited to a single city. Many police departments have joined a nationwide sharing network where Texas police can directly search Flock footage from Massachusetts, even tracking cross-state movements. Flock also sells AI surveillance cameras for tracking individuals, mobile surveillance trailers, and drones—the entire ecosystem keeps expanding.

Layered Problems: Technical Capability vs. Governance Gaps

As deployment numbers increase, problems surface:

1. Security vulnerabilities: Flock cameras run a modified version of Android, with footage wirelessly uploaded to the cloud for indexing. Multiple data breaches have been exposed.

2. Law enforcement abuse: There are no legal regulations governing police use of AI for reverse searches. Innocent drivers have received citations due to algorithmic errors (imprecise vehicle descriptions, poor lighting causing recognition failures), only to be exonerated later.

3. Public ignorance: Most drivers don't know their movements are being continuously recorded. Many cities have signed contracts without sufficient public debate.

4. Political dilemma: Some cities have wanted to abolish or restrict Flock systems. But the problem is—the harder it is to exit a contract, the harder it is to exit.

The Lock-In Trap: Why Can't We Just Remove It?

This is a textbook example of the "lock-in effect" in real life.

Once a city signs a Flock contract, it steps into a multi-layered trap:

First layer: Switching costs - Hardware costs of cameras (potentially tens of millions of dollars already invested) - Police departments have become accustomed to searching with Flock; switching systems requires retraining - Historical image databases exist only in Flock's system; migration is difficult

Second layer: Sunk cost of data - After 5, 10 years of accumulation, the image database becomes irreplaceable - Police increasingly rely on this massive archive to solve cases - Once deleted or switched systems, historical investigative capacity disappears

Third layer: Contract design - Flock contracts typically lock in long-term commitments (3-5 years or longer) - Early exit requires paying substantial penalties - Contract language is often vague; by the time cities discover surveillance capabilities exceed expectations, it's too late to turn back

Fourth layer: Institutional collusion - Police unions support continued use (improved case clearance rates, simplified work) - Local politicians fear being accused of "weakening law enforcement tools" - Once cross-state sharing networks are formed, a single city's attempt to exit faces opposition from other cities

Why This Is a "Principle" Rather Than "News"?

The significance of the Flock Safety story isn't that it's a bad company (though it's controversial), but that it exposes a universal phenomenon in tech governance:

Technology adoption decisions are often made before the technology's full capabilities are revealed. Initially, cities sign contracts for simple reasons—using AI to help police solve cases more efficiently. But after signing, system capabilities continuously upgrade (adding cross-state searches, personal tracking, drones), and risks continuously emerge. When cities want to reconsider, they discover themselves locked in by contracts.

This pattern repeats in any combination of "infrastructure + AI + long-term contract": - Prisons adopt facial recognition systems, later discovering accuracy bias against certain races - Hospitals use AI diagnostic systems, signing 10-year contracts then unable to upgrade - Governments procure certain surveillance software, discovering backdoors only after deep integration

The root problem isn't how bad Flock is, but rather: a democratic society shouldn't allow technology decisions to be locked in by contracts, preventing public will from changing course.

Three Paths Out (All Difficult)

1. Hard exit: Pay penalties, withdraw directly, discard all data. Nearly impossible for budget-constrained cities.

2. Negotiation: Demand Flock modify contract terms and reduce penalties. But Flock holds negotiating power ("you're already dependent on us"), limiting negotiation space.

3. Institutional innovation: Legislators establish "minimum lock-in clauses for technology contracts," requiring: - Contract length not exceeding 2-3 years - Early exit penalties not exceeding a certain percentage - Data portability rights (carry historical data when switching systems) - Periodic public review mechanisms (annual votes on continuation)

But such legislation typically lags technology deployment by 5-10 years; by then damage is done.

Takeaway

The Flock story reminds us: when government purchases large technology systems, it shouldn't just examine "current capabilities," but ask three questions:

1. What additional capabilities will this system have in 5 years? (Transparency of technology roadmap) 2. If public opinion changes, how quickly can we exit? (Unlock costs) 3. Who controls the data, and can we take it with us? (Data sovereignty)

The vaguer the answers, the higher the lock-in risk.