Expected practice
Identify original, derived, and synthetic objects separately. Document acquisition and transformations, including filtering and redaction. Preserve source links internally while removing credentials, session tokens, personal identifiers, and unnecessary private content from public exports.
Responsibility and boundaries
Real-person synthetic media, participant records, and harmful artifacts require particular care. Prefer minimized examples and aggregate findings when full release would expose people or facilitate abuse. Publication permission and consent are separate from possession of the data.
Application and current status
Explain restricted or unavailable data and offer safe reproduction material where possible. A digest supports byte comparison but does not grant a license or prove source authenticity. The datasets hub describes the reference downloads that actually exist.
Canonical reference: https://syntheticforensics.org/policies/data/
Report a technical correction