Rights & provenance
The part everyone else treats as a footnote.
Training-data deals fall apart over rights, not quality. These guides explain the landscape clearly — for buyers de-risking a purchase and owners understanding what they're signing. Kept current as the law moves.
New to the space? Start with the core trade-off — licensed vs. scraped training data — then go deep on the topic that decides your deal.
What "rights-cleared" training data actually means
The phrase, defined precisely.
Chain of title for training data
Why provenance is the deal-maker.
Voice and likeness consent for AI
Copyright alone does not reach it.
The NO FAKES Act, explained
The federal voice/likeness bill.
Training-data copyright lawsuits: the landscape
What the courts have actually said.
Is scraped data fair use for AI training?
The unsettled question.
Licensing vs. selling your data
For owners: keep ownership.
What diligence buyers actually run on training data
The checklist labs use.
NYT v. OpenAI, explained
The journalism fair-use test, unresolved.
Thomson Reuters v. Ross, explained
The first appellate fair-use test.
Getty v. Stability, explained
The UK ruling, read correctly.
UMG v. Anthropic, explained
The music-lyrics case, escalating.
The ELVIS Act and AI voice cloning
The first state AI-voice law.
GDPR and voice data
EU speakers change the analysis.
Biometric consent for speech data
Public audio is not consent.
Provenance records, explained
What the record actually contains.
Opt-in vs. opt-out for AI training
The two consent architectures.
Model-training rights in a licence
The anatomy of the grant clause.
The EU AI Act and training data
What Article 53 requires, and when.
The state publicity-law map for AI voice and likeness
The US patchwork, state by state.
Training-data warranties, explained
What suppliers promise, and what it is worth.
What happens when consent is withdrawn
Datasets, models, and the limits of undo.
Licensing music-adjacent audio for AI
Why singing is thicker than speech.
Related guides
More in the resources library: licensed vs. scraped — what's safe to train on, what consent for AI training looks like, and the US state voice & likeness law map.