---
title: "How Flock Cameras Are Using Artificial Intelligence"
description: "5 source-backed AI claims from How Flock Cameras Are Using Artificial Intelligence, with key statistics, context, and the original AI Adopters Club source."
url: "https://kbanc.com/claims-library/how-flock-cameras-are-using-artificial-intelligence"
source: "https://aiadopters.club/p/how-flock-cameras-are-using-artificial"
date: "2026-09-15"
topics: ["strategy", "business", "tools"]
generated: "2026-09-19"
---

# How Flock Cameras Are Using Artificial Intelligence

By Kamil Banc | September 15, 2026

## Claims

1. **Misquoted Statistic** (source summary): The widely cited 32.3 percent false positive rate came from a Futurism journalist's calculation, not the LAPD audit.
2. **Stale Hot Lists** (source summary): LAPD's audit attributed the 161 false alerts to stale hot list data, not camera misreads.
3. **Axon Cameras, Flock Software** (source summary): The footage analyzed came from roughly 1,500 Axon in-car cameras run through Flock's software.
4. **Patent Contradicts Public Claims** (source summary): A 2022 Flock patent covers classifying people by sex, race, and clothing, and matching faces.
5. **Denver's Reason for Switching** (direct quote): Denver's mayor said Axon doesn't maintain a national cross jurisdiction database the way Flock does.

## Evidence

### Quote
> "The moat and the liability are the same asset, looked at from two different angles." - Kamil Banc

### Key Statistics
- **32.3%**: False positive rate calculated by a Futurism journalist from 161 false alerts out of 498 total alerts; the figure never appears in the LAPD audit itself.
- **337**: Stolen cars recovered off Flock alerts over two months, according to the same LAPD audit, demonstrating the plate reader worked as intended.
- **$3,000 per camera per year**: Approximate all-in subscription cost for Flock hardware, sold on multi-year terms that auto-renew unless canceled.
- **83,000+ cameras**: Number of cameras reportedly touched by a single Texas sheriff's search logged as 'had an abortion, search for female,' spanning more than 6,800 Flock networks.

## Context
The author's methodology centers on tracing a viral statistic back to its primary source, then decomposing Flock's product into a chain of camera capture, model inference, database lookup, and human decision to isolate where errors actually occur. This practitioner lens treats the model as table stakes and redirects evaluation toward data ownership, contract terms, network reach, and switching costs. For buyers of AI tools, the actionable takeaway is to audit what the system is plugged into and who else can query it before purchase. For builders, the lesson runs in reverse: durable value accumulates in workflow lock-in, proprietary data, and distribution rather than in the trained model itself.

## Source
- Original: [How Flock Cameras Are Using Artificial Intelligence](https://aiadopters.club/p/how-flock-cameras-are-using-artificial)
- Cite: kbanc.com/claims-library/how-flock-cameras-are-using-artificial-intelligence

## Primary Evidence
- [LAPD’s own audit](https://www.oig.lacity.org/_files/ugd/b2dd23_781bdb1b27314ec4acc45f4b821320d8.pdf) (oig.lacity.org; supports claims 1, 2, 3)
- [a patent](https://patents.google.com/patent/US11416545B1/en) (patents.google.com; supports claim 4)
- [Denver’s mayor said Axon doesn’t maintain a national cross jurisdiction database the way Flock does](https://www.denverpost.com/2026/02/24/denver-flock-license-plate-cameras-contract-axon/) (denverpost.com; supports claim 5)