A printed pattern that makes you invisible to AI surveillance

A new computer-generated pattern printed on a hoodie can trick AI surveillance cameras into ignoring you.

Security researcher Bill Swearingen spent a year running 31 million automated tests to create noRecognition. The project uses a self-training AI model that learned how to "paint" patterns that confuse detection software. The prints stop cameras from triggering alerts. They beat 11 algorithms used by Flock license plate readers, Axon body cams, and Clearview AI. At Def Con, Swearingen wrapped a 2009 Toyota Yaris in the pattern and successfully tricked a Flock camera in a live test.

Why it matters: Automated surveillance works by using software to pull needles out of haystacks. These patterns do not block the camera from recording video footage. Instead, they scramble the software's ability to identify objects, turning you back into the haystack so you can opt out of algorithmic tracking in public.

Swearingen is running a crowdsourcing campaign to fund early merchandise featuring the patterns, from T-shirts to vehicle skins. The project is early, but the model creates updated pattern recipes every minute.

Finally, a fashion choice built specifically to annoy an algorithm.