🤖👕 A Berlin artist created a T-shirt designed to make you “invisible” to AI cameras.
At first glance, it looks like a loud Hawaiian shirt. But its pattern isn’t designed to fool humans—it’s designed to confuse computer vision algorithms.
The artist tested different patterns against YOLO, an object-detection system, looking for designs capable of interfering with how the model identifies the human silhouette.
Under certain conditions, the result can be striking: the camera detects other people while the algorithm may fail to classify the person wearing the shirt as a human.
But the shirt doesn’t make anyone literally invisible. It is an example of an adversarial attack against specific AI vision systems, and it doesn’t work against every model or under every condition.
The same artist previously made headlines in 2020 after walking through Berlin with 99 smartphones in a small wagon, creating a simulated traffic jam on Google Maps.
And that raises a bigger question.
The experiment isn’t really about a T-shirt.
It’s about what happens when AI systems are increasingly used to identify and monitor people, while the algorithms making those decisions remain closed, proprietary and difficult to independently audit.
The shirt is the experiment. The real debate is what happens when the algorithm watching you cannot be watched by anyone. 👁️🤖
