★ Independent, reader-supported & ad-free · Watching the watchers in all 50 states ★ Support Us
Watch · Our Take

Should police be able to tap into facial recognition software?

12 News · 8 years ago

The short version: 12 News revisits Amazon's Rekognition facial ID tool, a reminder that plate readers and face scanners are building the same surveillance net.

Our Take

Credit to 12 News for putting a plain question to viewers: should cops get to run your face through Amazon's Rekognition software? It sounds like a narrow tech-policy debate, but it's really the same fight we track every day with license plate readers — should private companies get to sell governments the ability to identify and track ordinary people, with little public input and even less oversight?

Facial recognition and ALPR cameras are two branches of the same tree. Both convert a public space into a biometric or vehicular checkpoint. Both get pitched to city councils as a simple crime-fighting upgrade, and both quietly create a searchable historical record of where you've been and who you're with. Amazon's Rekognition drew backlash years ago for misidentifying people, particularly people of color, at higher rates — a flaw that matters a lot when the output feeds directly into an arrest. Flock's plate cameras have their own error and misuse problems, but the underlying issue is identical: a private vendor builds the surveillance infrastructure, then hands police the keys with minimal transparency about audits, retention, or who else gets access.

The lesson from the Rekognition fight is that public pressure works — cities banned it, Amazon eventually paused police sales. That only happens when people know the tech exists in their town. Check our camera map to see what's already watching your streets, and use our take-action guide to push back before facial recognition and ALPR data get quietly linked together.

This is DeFlock The USA’s original commentary. The video above is the work of 12 News, published on YouTube — full credit to the creator.