No. L22025–26Image protection for artists
Working buildGhostprint
Adds a perturbation people cannot see to an image so that AI models misread it, to protect artists’ work from unauthorized training. Tested against ResNet, ViT, CLIP and the Stable Diffusion VAE within a set perceptual budget.
The current version optimizes against an ensemble of vision models with momentum and random transformations, so the protection survives resizing and compression. It also ships as an MCP server so agents can apply it in a pipeline.