Flock Ships American Surveillance Footage to Offshore Contractors
Contractors in the Philippines are labeling faces and tracking vehicles through 80,000+ Flock cameras. Local police installed them; Flock ships the data overseas.
Right now, contractors in the Philippines are watching American streets through more than 80,000 surveillance cameras. They’re labeling faces, tracking vehicles, and categorizing the sounds of children screaming. All through Flock Safety’s annotation platform. Local police departments installed these devices. Flock ships the data overseas.
Ever wonder how Flock’s AI cameras distinguish between different types of vehicles? How they differentiate between children screaming and adults screaming? It’s not magic — it’s a process called “data annotation” which uses an army of offshore gig workers to manually pore over images, videos, and sound files.
This post will be heavy on images and other data pulled directly from Flock’s annotation app.
They largely speak for themselves.
The “Springbank” Annotation Tool[1]

Flock created this tool so its Upwork team, mostly foreign gig workers, can review images and audio files taken by the 80,000+ cameras and microphones liberally sprinkled throughout American cities.
To put it in Flock’s own words:
Work is assigned to annotators with varying priority. For annotators that work on multiple different projects, this priority can change frequently. It is best to follow the outlined workflow to ensure the highest priority work is getting completed first, in addition to checking UpWork.
The company tracks metrics on its Upwork contractors who work to assign labels to ALPR images and sound files:

Flock’s performance dashboard shows its top annotators include workers we’ve confirmed are based in the Philippines, along with contractors whose names suggest Eastern European origins.
Do the math: 80,000 cameras generating thousands of images daily, processed by contractors bidding the lowest hourly rates. That’s potentially millions of American surveillance moments flowing through laptops in Manila and god-knows-where-else. No security clearances. No background checks. No oversight.


Flock CEO Garrett Langley tells city councils his cameras ‘don’t identify or track people.’ But Flock’s own training materials instruct offshore contractors to do exactly that — label every person.
The company built an entire workflow around human identification while swearing to American communities they’d never do it.

The campaigns also include audio annotation campaigns. While Flock publicly markets Raven for gunshot detection, the annotation categories reveal far broader audio surveillance.

Flock’s microphones aren’t limited to gunshot detection; they record conversations near ATMs, discussions outside abortion clinics, prayers at mosque entrances.
This annotation goes far beyond gunshot detection. Training documentation has detailed instructions and audio samples for annotators to label and categorize:
- Glass shattering
- Tire screeching
- Engine revving
- Gunshot/firework/explosion
- Car wreck
- Industrial banging
- Metal grinding
- Thunder
- Alarm
- Screaming adult
- Sawing
- Sawing vs. Leaf blower
- Background noise
Flock even remarks to its team of annotators that:
Given how hard it can be to distinguish between children and adult screaming, please make use of the certainty dropdown shown on the right.
Somewhere in Asia or Eastern Europe, someone is listening to American children in distress, cataloguing their terror for a few dollars an hour.
The (National) Security Issues
This is the surveillance equivalent of letting Huawei run our 5G networks. Foreign nationals now possess detailed movement patterns of FBI agents driving to work, military personnel visiting family, and intelligence officers meeting sources. In any other context, we’d call this intelligence collection. Here, we call it ‘data annotation.’
There is no oversight on these apps, or on these workers. Flock’s contract permits this use of the data, and Flock’s customers assign irrevocable, worldwide licenses to Flock specifically for this purpose.
The apps do not have to conform to any security standards, and, while there may be contractual terms, it is unlikely that Flock effectively oversees data security for data in the hands of its contractors.
Even if it wanted to, Flock would have very limited ability to enforce its employment agreements in many of these countries.
While Senator Wyden writes stern letters about police departments forgetting to enable two-factor authentication, Flock’s contract explicitly authorizes this overseas data hemorrhage. Every city council that approved these cameras blessed this arrangement. They traded your privacy for crime statistics, then acted shocked when the bill came due.
Flock Safety turned American streets into a foreign intelligence goldmine. Your mayor signed the contract. Your police chief installed the cameras. And tonight, someone eight time zones away will watch your neighbors come home, label their faces, and move on to the next task.
Request your city’s Flock contract. Search for Section 4.1, and the ‘worldwide license’ clause:
Customer hereby grants to Flock a limited, non-exclusive, royalty-free, irrevocable, worldwide license to use the Customer Data and perform all acts as may be necessary for Flock to provide the Flock Services to Customer.
And look for Section 4.3, which promises that “These images [for machine learning] are never sold or shared with third parties.” Apparently, foreign contractors reviewing your family’s daily movements don’t count as ‘third parties’ in Flock’s dictionary.
Then ask your mayor why Bulgarian freelancers need to analyze your daughter’s walk to school.
[!note] The information in this post was sourced exclusively from documents posted publicly online by Flock. None of the contractors leaked information, and no unauthorized systems access was involved.
Flock’s internal systems are all named after various alcoholic beverages. From “Springbank” handling audio annotation to “Minderaser” handling incident reports and even the flagship FlockOS “Strawberry Soju.” ↩︎