Definition
Demands for disclosure of datasets, model weights, and internal AI development practices used to train generative models — increasingly central in copyright litigation and regulatory debates.
Midjourney Discovery Motion (July 2026)
midjourney seeks Hollywood studios’ training datasets and model weights to support fair-use and unclean-hands defenses. Magistrate limited scope to consumer-facing AI; broader discovery pending (2026-07-06-midjourney-hollywood-variety).
Broader Context
- Plaintiffs suing over training data may face reciprocal disclosure obligations
- Internal studio AI for storyboarding could establish industry custom of training on unlicensed content
- Precedent affects all generative AI copyright cases