Berlin-based artist Simon Weckert has created a Digital Camouflage shirt designed to confuse artificial intelligence surveillance cameras into failing to detect human wearers.
The garment features a swirling pattern of bright colours, gradients and abstract shapes. According to Mr Weckert, the outlandish design confuses computer-vision systems so severely that the algorithms fail to recognize the wearer as a human being.
The project exploits a technique known in machine learning as an adversarial attack, where images are deliberately altered to fool automated vision models. While human observers can easily spot the person wearing the shirt, surveillance software sees nothing.
"Digital Camouflage is not a tool for evading the police, and it is not a promise of anonymity," Mr Weckert said. "No pattern can promise to defeat every system under every condition - but worn in public, it is a visible signal that being read by machines is not something everyone consents to."
Response to Berlin police testing
Mr Weckert developed the garment in response to the rollout of AI-assisted video surveillance at Kottbusser Tor in Berlin. Local police have been testing automated systems at the location to monitor and analyse human behaviour in public spaces.
Berlin is the capital of Germany, and Kottbusser Tor is a major urban square and railway interchange in the city's Kreuzberg district. The area is known for heavy pedestrian traffic, making it a focal point for public safety initiatives and video monitoring trials.
The artist argued that automated surveillance raises difficult questions about personal privacy and the growing role of automated monitoring in daily life. Demonstration videos released alongside the project show AI surveillance software drawing bounding boxes around every passer-by in a public square except the individual wearing the shirt.
"The algorithm no longer registers a person," Mr Weckert said. "The wearer becomes, in effect, invisible to the machine - present to every human eye, absent to the model."
"When systems that cannot be publicly tested can be undone by a piece of fabric, how much trust should we place in them before we hand them the public square - in Berlin, or anywhere else?" Mr Weckert said.
Commercial release and practical limitations
The shirt is currently on sale for $75.60 (£55.65). It is offered alongside other items designed by Mr Weckert, including a handheld Bluetooth jammer.
Computer-vision technology relies on deep-learning neural networks trained on vast datasets of imagery to recognize human shapes and facial features. Adversarial attacks target these neural networks by introducing specific visual noise or conflicting patterns that cause the software to misclassify objects.
Although the shirt may hide a wearer from automated cameras, it presents an obvious practical limitation. The garment's loud psychedelic design is highly eye-catching to human observers, making wearers among the most noticeable individuals on the street rather than helping them blend into a crowd.
AI-powered surveillance is also becoming increasingly common across the UK, where police forces use live facial-recognition cameras in public spaces to spot wanted suspects. Supporters state that the technology aids crime prevention, while critics express concern over privacy rights and the automated tracking of innocent citizens.
A Bluetooth jammer is a device designed to disrupt short-range wireless communications by broadcasting signals on the 2.4 GHz radio spectrum. Such equipment prevents nearby Bluetooth-enabled electronics from establishing or maintaining stable data connections.
