London’s Metropolitan Police used live facial recognition to make over 1,000 arrests since early 2024. That means every face in a crowd is being scanned whether you consented or not. Against that backdrop, a handful of designers are now selling clothing engineered to make surveillance apps and cameras see something other than a human being. The pitch is seductive. The reality is more complicated.
Adversarial Design: Dressing to Confuse the Algorithm
A growing category of garments uses algorithmic patterns to exploit weaknesses in computer vision systems.
Cap_able, an Italy-based startup, knits adversarial patterns directly into sweaters and pants. The arrangements of shapes and repeated motifs reportedly cause object recognition software to misclassify the wearer as an animal — a giraffe, a dog, according to CNN coverage. To your eyes, it’s a bold textile print. To the camera, it’s a zoo exhibit. The technology behind this is adversarial machine learning: deliberately crafted inputs designed to trip up the pattern-recognition systems that underpin most modern facial recognition pipelines.
The product range extends beyond knitwear. Several companies offer distinct approaches:
- Urban Privacy markets a coat with integrated infrared lighting, invisible to the human eye, purportedly designed to overwhelm surveillance cameras.
- Reflectacles and Sunphey Optical sell eyewear with reflective coatings and infrared-blocking features targeting the eye-detection step that most recognition pipelines depend on.
This is the closest real life has gotten to a Mr. Robot wardrobe drop — except you can actually buy it. Designers have suggested to outlets including The Guardian that privacy could become a defining fashion concern, positioning this niche as something more than novelty.
The Arms Race Problem No Jacket Can Solve
Adversarial garments work against specific models, but surveillance software evolves faster than fashion cycles.
Here’s where honesty matters more than hype. Mozilla Foundation reporting found that adversarial garments can defeat specific recognition models, but “widely-varying and rapidly-evolving algorithms” quickly erode that effectiveness. Your jacket is static. The surveillance model gets retrained. You’re wearing last season’s camouflage against this season’s cameras — roughly as useful as trying to block every spoiler in a post-Succession group chat.
Critics argue live facial recognition treats every passer-by as a potential suspect, according to Reuters — a framing that helps explain why these products have a political market beyond novelty. Reports of apps secretly tracking users have only deepened that concern.
Right now, the buyers skew toward activists, artists, and privacy-focused early adopters in high-surveillance environments rather than everyday commuters. If you’re a general consumer outside a major surveilled city, lower-tech options may serve you just as well:
- hats
- sunglasses
- masks
- route planning using tools like the Electronic Frontier Foundation’s Atlas of Surveillance
The category grew from art projects and small startups, not mass-market labels, and that origin still shapes who reaches for it first. Wearing adversarial fashion turns privacy from an abstract policy debate into something visible on your body. That statement might matter more than the pattern itself.
