Run a local Large Language model
The short version: A video on running large language models locally is a reminder that keeping compute off corporate and government servers is a privacy move worth applying to surveillance tech too.
Our Take
This one isn't about license plate readers directly, but the underlying principle lines up with everything we cover. The channel Digital ID, Mass Surveillance and Digital Currency walks through how to run a large language model on your own hardware instead of funneling every query through a corporate cloud. Credit to the creator for making that process approachable — most people assume AI has to mean handing your data to a tech giant, and that assumption is exactly how mass surveillance normalizes itself.
The same logic applies to automated license plate readers. Flock Safety and similar vendors build business models on centralizing data that used to be nobody's business — where you drove, when, how often — and storing it on servers you don't control, shared with police departments and sometimes federal agencies with little public oversight. Whether it's an LLM or an ALPR network, the question is the same: who holds the data, who can query it, and what happens when that system gets subpoenaed, hacked, or quietly expanded.
Keeping your own compute local is one small act of resistance. Knowing where the cameras watching your street actually are is another. Check our camera map to see what's been documented near you, and visit our take-action page if you want to push back on these systems in your own community.
This is DeFlock The USA’s original commentary. The video above is the work of Digital ID, Mass Surveillance and Digital Currency, published on YouTube — full credit to the creator.