A practical provenance question for this list: if an image producer signs a manifest containing a hash of the output and the claimed creation steps, a verifier can check integrity and the signing key, but not whether the depicted event occurred or whether the signer was truthful. Key compromise, metadata stripping, and later editing complicate the chain further. What verification practices make this distinction clear to readers without turning signatures into a false authenticity badge? I would be interested in approaches that avoid linking unrelated uses of the same signing key or revealing prompts and private source material. Disclosure: I work on https://realisticaiimagegenerator.online/ and used AI assistance to draft this question. The example is about synthetic image workflows, not a claim that any particular picture is authentic. Thank you.