AI-Generated Pattern Evades Flock Surveillance Cameras
Bill Swearingen’s noRecognition project uses adversarial patterns to defeat AI object detection, including Flock, Axon, and Clearview. At Def Con, a wrapped Toyota Yaris went undetected by a Flock camera. The project aims to let people opt out of tracking and is crowdfunding merchandise.
Quick Take
Patterns defeated all 11 open-source detection algorithms tested, including Flock and Clearview.
Public Def Con demo showed a wrapped Toyota Yaris evading a Flock camera.
Project uses reinforcement learning to generate new patterns every minute.
Swearingen calls privacy a fundamental right; crowdfunding merchandise now.
Market Impact Analysis
NeutralNo crypto-specific market impact; this is a privacy/surveillance technology story unrelated to digital assets.
Speculation Analysis
Key Takeaways
- Adversarial pattern defeated all 11 open-source detection algorithms tested, including Flock license plate readers, Axon body cameras, and Clearview AI.
- Public Def Con demo showed a wrapped 2009 Toyota Yaris evading a Flock camera while normal video recording continued.
- Pattern uses reinforcement learning to generate new designs every minute, with strongest versions kept offline to prevent counter-training.
- Swearingen calls privacy a fundamental right, launching crowdfunding for pattern merchandise and apparel.
What Happened
Bill Swearingen's noRecognition project unveiled adversarial patterns that blind AI object-detection software while leaving video footage intact. At Def Con in Las Vegas, a 2009 Toyota Yaris wrapped in one of these designs rolled past a Flock surveillance camera. The camera recorded normally, but its automated system never logged the vehicle. Swearingen developed the approach after wanting to attend a protest without being tracked. The patterns trick computer vision models into ignoring what they cover, effectively making people and vehicles invisible to automated logging systems. The project launched a crowdfunding campaign to bring the patterns to apparel and vehicle skins.
The Numbers
The system defeated all 11 open-source detection algorithms Swearingen tested. That includes software behind Flock license plate readers, Axon body cameras, and Clearview AI facial recognition. Training involved about 31 million test iterations. The reinforcement learning model grades its own output, adjusting until detectors fail to classify the target. Fresh patterns emerge every minute, while the strongest versions stay offline to prevent camera vendors from training against them. The public Def Con test used a 2009 Toyota Yaris, and Donut Media will release demo footage in the coming weeks.
Why It Happened
Swearingen wanted to attend a protest but worried cameras would log everyone present. That led him to explore adversarial machine learning—designing visual noise that exploits how computer vision models process images. The patterns do not disable cameras; they only disrupt the object-detection layer that classifies vehicles, faces, and plates. Because classifiers learn statistical features rather than seeing like humans, carefully engineered patterns can make a car invisible to AI while remaining obvious to people. Growing deployment of Flock cameras across America created urgency for privacy tools that let individuals opt out of automated tracking.
Broader Impact
This work highlights a structural weakness in surveillance AI: any detector trained on known features can be fooled by adversarial inputs. It echoes grassroots tactics like traffic cones on robotaxis, but moves beyond physical obstruction to algorithmic evasion. If adopted widely, such patterns could reduce the effectiveness of mass surveillance systems, forcing vendors into an arms race of retraining. The project also reframes privacy as a technical opt-out, not just a policy debate.
What to Watch Next
- Donut Media video release in coming weeks will show the Def Con demo in detail, including any limitations like wheel detection.
- Swearingen plans to refine patterns for apparel and vehicle skins, and crowdfunding progress may reveal demand for privacy-focused merchandise.
- Camera vendors may respond with model updates or attempts to train against public patterns; watch whether the strongest patterns remain effective.
This article is for informational purposes only and does not constitute financial advice.
Always late to trends?
Join for the latest news, insights & more.
Disclaimer: Bytewit is an independent media outlet that delivers news, research, and data.
© 2026 Bytewit. All Rights Reserved. This article is for informational purposes only.
