Pre-Publication V2. The process of obscuring, removing, or transforming information about the origin, ownership, licensing status, or rights associated with data or content in a way that makes its provenance or permitted use more difficult to identify. This may occur when content is processed, modified, or incorporated into an intermediary dataset or AI-generated output and subsequently used as source or training material without clear information about the underlying data or associated rights.
Deliberation Summary:
Suggestion to add clarity in the definition; example offered of a developer training a model on generated images sourced from other systems whose own training material was obtained without permission, while presenting its own model as ethically sourced; the effect is that the origin of the underlying material becomes difficult or impossible to trace unless an output is unmistakable.
Pre-Publication V1. The process of obscuring, removing, or transforming information about the origin, ownership, licensing status, or rights associated with data or content in a way that makes its provenance or permitted use more difficult to identify.