Generative and agentic AI has made one question central to entertainment and sports law: when a model is trained on a real person's voice, face or style, who controls the output? In 2026, that question is being answered in contracts long before it is answered in court.
Training data is a licensing problem
If a company ingests an artist's catalog, an athlete's footage or a creator's likeness to train a model, that ingestion should be licensed — with defined scope, compensation and the right to withdraw. Rights holders who address training data at the contract stage keep leverage that is nearly impossible to claw back afterward.
Protecting the digital twin
- Consent: explicit, revocable permission to create or use a synthetic likeness.
- Attribution and control: who may deploy the twin, in what contexts, and who approves.
- No-training clauses: barring use of your data to train systems without permission.
PARRON LAW advises artists, athletes, studios and technology companies on model licensing, training-data rights and synthetic-media risk. If AI touches your name, image or likeness, book a consultation to get the structure right before the product ships.