Qatar has ordered 15,000 new surveillance cameras for the World Cup, and now has a total of 250,000 cameras with built-in facial recognition monitoring visitors day and night.

This isn’t just problematic because Qatar has a problematic relationship with human rights and offers very little legal certainty. It’s also deeply problematic from a privacy standpoint because there is no transparency about what data is collected, what it is used for, and who subsequently gets access to your most valuable identifier: Your face.

Insight:

You can prevent facial recognition with costumes and make-up based on geometric disguises. But these need to be so radical that the very things that fool the algorithm look conspicuous to the human eye. You might hide from the AI; but you will attract even more attention from guards and observers!

Unless, of course, you can make the disguise itself blend in with your surroundings...

IDEA

Camouflags — Anti-surveillance patterns fashioned like football fans' face paint.

The series of designs are visually based on the flags of the participating countries, but the patterns are developed, based on, and tested against facial recognition technology's blind spots.

We developed the patterns by turning artificial intelligence against itself.

We fed the midjourney algorithm images of verified anti-surveillance patterns and asked it to imagine them as flags of different countries.

We tested the output against various open-source machine learning models (e.g. Google's FaceNet and DeepFace), selected the best performing ones and fed them back again. We repeated this looping feedback process until the patterns broke the algorithm.

The result was launched as both an inspiring photo series, gonzo journalistic experiment and interactive tutorial through SparkAR filters in Instagram — and kicked off a global conversation about data security, surveillance and facial recognition.