According to Music Business Worldwide, Google is pushing a legal theory that would treat the ingestion of copyrighted works for AI training as categorically fair use, shifting all infringement scrutiny to what models generate rather than what they consume. If regulators or courts adopt that framing, it would sharply narrow the leverage rights holders and data licensors currently have over scraping and dataset sourcing.
For the data economy, the input/output distinction is the ballgame: it determines whether training-data provenance remains a compliance issue at all, or whether licensing becomes relevant only after the fact, at the point of generation.
Google says AI training is fair use and copyright should be policed on outputs, not inputs
