Designer creates clothing pattern that blocks artificial intelligence object recognition

Artist Simon Weckert has created a shirt featuring a specialized graphic pattern designed to baffle artificial intelligence vision systems. As detailed on Weckert's project website, the clothing relies on adversarial patterns that scramble the specific visual features object detection software uses to spot people. To a standard computer vision model, a person wearing the garment simply disappears from the scan.

The project comes as cities expand automated monitoring, such as new behavior tracking cameras installed in Berlin's Kottbusser Tor area. These tools attempt to classify everyday movements as normal or suspicious. Weckert notes that city officials frequently deploy secret proprietary software that cannot be audited by the public, even though basic open source models can be tricked by simple textiles.

The initiative is intended as a privacy statement rather than a foolproof tool for police evasion, as different camera networks rely on varied detection methods. Proceeds from sales of the garment are being donated to civil rights groups that campaign against pervasive public surveillance.

This creative pushback highlights a growing gap between civic privacy and machine monitoring. As computer vision becomes cheaper to install on public streets, society will have to decide whether citizens should hold the right to remain unparsed by algorithms while simply walking down the road.