How it works: Untrainable applies invisible adversarial noise to your image based on the Nightshade research. When AI models train on poisoned images, they learn corrupted associations. Your original image is never sent anywhere — all processing runs locally in this tab.