6.1 The Training Data Dividend
The Training Data Dividend is the compensation owed to a rights holder whose creative work has been used as AI training data. The CIP approach is pay-per-contribution-path: each creator is compensated based on their specific contribution to the training corpus, not just dataset inclusion.
This requires CDR infrastructure — the Rights Registry tracks which assets went into which training corpus, enabling asset-level compensation rather than dataset-level approximation.
6.2 The revenue waterfall
When a licensed training event occurs, compensation flows through the revenue waterfall:
- Platform payment — the AI company pays a Training Data Dividend to the rights pool
- Pool distribution — distributed to individual rights holders based on CDR provenance records
- Rights holder receipt — the creator or their authorised representative receives payment
- Reporting — automated reporting through the Rights Registry records the payment event
For creators working through agencies, the waterfall may include an intermediary step where the agency receives payment and distributes to the creator under the terms of their representation agreement.
6.3 NILP Downstream Obligations as royalty triggers
When AI-generated content involving a person's identity is commercially distributed, the NILP Downstream Obligation creates a royalty trigger that runs from the AI company through to the rights holder — regardless of how many commercial intermediaries were involved.
A brand that used an AI platform to generate a voice advertisement using a synthesised celebrity voice owes a NILP Downstream Obligation to that celebrity — even if the brand paid the AI platform and none of those payments reached the celebrity.
6.4 Mechanical licences in AI contexts
A mechanical licence is the traditional royalty mechanism for reproducing a musical composition. In AI contexts, mechanical licences may apply where AI-generated music is based on a specific copyrighted composition or where an AI output is determined to be a derivative work.
Summary
Key Takeaways
- Training Data Dividend: compensation per contribution path, not per dataset inclusion
- Revenue waterfall: AI company → pool → rights holders via CDR provenance records
- NILP Downstream Obligations create royalty triggers for commercial identity use in AI outputs
- CDR registration is the practical prerequisite for receiving Training Data Dividend payments
Self-check
Check Your Understanding
- What is the Training Data Dividend?
- What compensation model does the CIP framework use for the Training Data Dividend?
- In the CIP revenue waterfall, what enables asset-level compensation to individual rights holders?
- A brand uses an AI platform to generate advertising with a synthesised celebrity voice. Who bears the NILP Downstream Obligation?
- What is the practical prerequisite for receiving Training Data Dividend payments?