Studios and brands are not debating whether generative AI enters the pipeline. It's already in it: concept passes, temp frames, pitch boards, localization. The debate is whether anyone will be able to tell, a year from now, where a given image came from and whether it was cleared.
That is a provenance problem, and provenance doesn't retrofit. It has to be applied at the moment of creation, by the system doing the creating.
Our approach in Asset Foundry is provenance by default. Visible watermarking is available on video, image and document outputs, with position, opacity and tiling under your control. Image outputs can carry C2PA Content Credentials, the open provenance standard, including an assertion that the image was AI-generated. Generation runs behind one governed API where every job is logged, and models can be pinned to the providers you approve, or kept local.
The result is simple to state: any frame that leaves the system can explain itself. Where it came from, what made it, whether a person approved it. When legal asks, the answer is in the file, not in someone's memory.
Generative capability is easy to buy. Governed generative capability is what studios can actually use.