Berlin-based artist Simon Weckert has created a button-up shirt called Digital Camouflage that AI surveillance cameras reportedly fail to register as a human being. The garment was designed as a direct response to the AI-powered video system switching on this month at the city's Kottbusser Tor transit hub.

How can a shirt protect you from surveillance?

Weckert, an artist from Berlin, released Digital Camouflage, a shirt that can reportedly reduce an object-detection system's confidence in recognizing a human being. According to Weckert, these systems never actually learned what a person is — they only learned what people tend to look like across millions of training photos: edges and texture in the early layers, then the head-and-shoulders outline and limb proportions deeper in.

Digital Camouflage exploits that training method with a loud tangle of pink, orange, and green shapes. Weckert explained that the saturated colors light up a network's first layers, while the overlapping forms distort the body's outline to the point where the detector cannot stitch the pieces back into a single figure. To a human eye, the shirt looks like a hypnotic Hawaiian print; to surveillance models, it reads as nothing.

Weckert built the pattern through an "adversarial loop": he would generate a candidate print, show it to a detector, measure how confident the software still was that it saw a person, then adjust and repeat until he achieved satisfactory results. He reportedly used a generative method labeled TC-EGA (Toroidal-Cropping-based Expandable Generative Attack) to run those rounds, and tested the results against YOLO, an open-source, real-time detection system built on the same class of technology behind many commercial surveillance products.

The approach builds on a well-established body of computer-vision research: adversarial examples — small, often imperceptible or patterned perturbations that cause neural networks to misclassify objects — have been documented in academic literature since at least 2013, and researchers have previously applied the concept to physical garments, from adversarial eyeglass frames to patterned patches meant to confuse detectors. Weckert's contribution is to turn that research into a wearable consumer product aimed squarely at a public deployment.

Weckert clarified, however, that the shirt was not engineered to defeat one particular government camera, but to raise the question of how much trust a system deserves if cheap fabric can defeat it. On his own site, he stresses that the shirt is "not a promise of anonymity" and makes no claim about any specific deployment, precisely because outsiders cannot test those systems. The garment sells for roughly $88, with a cut of each sale going to digital civil-rights groups.

Why are people suddenly looking for protection from surveillance?

Weckert's website explains that cameras — which he calls behavior scanners — are set to be installed at Kottbusser Tor, a busy U-Bahn station in Berlin's Kreuzberg district. The software is meant to sort passers-by into "normal" and "suspicious." He believes three more Berlin locations are next in line.

The deployment lands in a Germany with a strong historical aversion to expansive monitoring, rooted in the surveillance practices of both the Nazi era and East Germany's Stasi, and Kottbusser Tor in particular has seen repeated privacy protests over video monitoring proposals in recent years. It also comes as the European Union's AI Act — the bloc's wide-ranging law classifying and restricting AI systems by risk, including those used in public spaces — is being phased in.

In the United States, Texas Governor Greg Abbott ordered state agencies on a Thursday night to stop funding Flock Safety, whose license-plate readers number more than 120,000 across 49 states. Republican Senator Josh Hawley also opened an investigation into how the company handles captured data.

Flock Safety's license-plate readers have drawn significant backlash following reports of law enforcement officers severely abusing the system for personal purposes, such as tracking ex-girlfriends, spouses, and romantic rivals. In one extreme case, a Kentucky officer accessed the Flock system approximately 2,000 times to monitor his former partner's movements. Another officer in Georgia, who was also a police chief, conducted over 600 searches of his ex-girlfriend and her daughter before later taking his own life while facing charges for stalking and misuse of police equipment.

The shirt is not Weckert's first attempt at pushing back against these systems or sending a message. Back in 2020, he dragged a wagon of 99 phones through Berlin to fake a Google Maps traffic jam, demonstrating that AI-driven systems are easily manipulated.