Berlin artist designs shirt to disrupt AI object detection

What the shirt does

According to YourStory, Berlin-based artist and technologist Simon Weckert has created a garment called Digital Camouflage—a Hawaiian-style shirt printed with bright, abstract patterns. The patterns are designed to interfere with computer-vision systems that perform object detection.

The shirt does not render the wearer invisible to human observers or standard cameras. Instead, it targets the AI software analyzing the video feed, introducing visual noise that disrupts the model’s ability to consistently identify a person.

Weckert’s demonstrations, cited by YourStory, use the open-source YOLO object-detection system. In these tests, the system identifies other pedestrians as people but fails to consistently recognize the wearer of the shirt.

How it works

AI object-detection models learn to recognize patterns such as body outlines, shapes, and edges. Digital Camouflage uses an adversarial attack—a technique that introduces subtle visual information designed to confuse the model’s learned patterns.

The effectiveness of the shirt depends on multiple factors, including the specific AI model, camera angle, distance, lighting, and how the fabric drapes on the wearer. YourStory notes that results can vary across different detection systems.

Why the project exists

Weckert developed the shirt in response to plans for AI-powered video surveillance at Berlin’s Kottbusser Tor, a busy public square. According to YourStory, he frames the project as an artistic intervention rather than a commercial privacy tool.

The goal is to expose a vulnerability in machine vision: AI systems can appear highly capable but remain sensitive to small, imperceptible perturbations. The shirt makes this technical limitation tangible by turning it into something wearable.

Limitations

The shirt is not a foolproof privacy solution. YourStory reports that its effectiveness varies with changes in camera angle, distance, lighting, and movement. It does not block facial recognition or sophisticated tracking systems.

Weckert does not position the shirt as a reliable defense against law enforcement cameras. Its role is primarily artistic and conceptual, aimed at sparking discussion about AI surveillance rather than providing guaranteed protection.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

Simon Weckert’s project treats adversarial patterns as wearable critique rather than commercial product. For founders and operators, the shirt underscores two asymmetries that matter beyond art:

1. Low-cost evasion versus high-cost detection – A T-shirt pattern can degrade the performance of widely deployed object-detection models, yet closing that vulnerability may require costly retraining or architectural changes. If the vulnerability scales as these models are embedded in more public infrastructure (traffic, transit, retail), companies building surveillance tools or downstream applications may face recurring patch-and-update cycles that hurt gross margins.

2. Transparency risks – The project makes a latent technical flaw visible to the public. That visibility can become brand risk: customers who see a viral video of a person ‘disappearing’ from a security feed may question the reliability of the entire platform. Founders in the B2B computer-vision space should consider whether their contracts include performance SLAs for adversarial robustness and who bears the cost of remediation when patterns like Digital Camouflage emerge.