Privacy boundaries for image-reference workflows
When a creative tool accepts a reference image, the file can carry much more information than the visible subject: embedded metadata, faces in the background, filenames, and details that identify a location. A privacy review should consider each stage separately: browser preview, upload, transient processing, retained outputs, and any logs or backups. I work on AI Anime Generator (https://ai-anime.io/), and I am interested in practical ways to explain these boundaries to users without implying that a browser-based workflow is automatically private. This message was drafted with AI assistance. A few design questions for people who have reviewed similar systems: how do you communicate which metadata is stripped, how long original uploads remain accessible, and whether deletion includes derived copies? What evidence or independent tests would make those statements verifiable? Consent from people depicted in reference images and clear choices about sharing outputs seem as important as a short retention notice. I would welcome pointers to public threat-model templates or useful failure cases, especially for accidental inclusion of private details. I am not claiming a specific retention period or security guarantee for any product here.
participants (1)
-
support@ai-anime.io