Midjourney's defense in the lawsuit filed by several film studios rests on a simple question: and you, how do you use artificial intelligence? The company behind the eponymous image generator has asked a US federal judge to compel the producers to disclose their internal use of AI models. The news, reported by Variety, marks a turning point in a case that accuses the company of training its models on copyrighted works without permission.

At stake is transparency. Midjourney does not deny using large datasets for training, but it turns the tables: if the problem is the use of protected data, then the studios should also prove they haven't done the same. The move feels like a strategic escalation, because forcing major studios to expose their AI pipelines could reveal internal practices that have so far been opaque. It is no secret that film productions are exploring generative models for concept art, storyboards, and even visual effects.

This matter fits into a broader debate about copyright and artificial intelligence. While tech companies push for ever more powerful models, rights holders demand clear rules. The Midjourney case could set a precedent: if a court orders a comprehensive discovery of data use from both sides, it would establish new standards of due diligence for any organization that uses models trained on public data.

For those evaluating on-premise deployment of LLMs or image generators, the issue has practical repercussions. Control over data provenance and training methods becomes a legal risk factor that cannot be overlooked. In regulated environments, the ability to trace datasets and ensure compliance is already a requirement. Midjourney's request, if granted, could accelerate the adoption of audit tools for self-hosted models.

In this context, dataset management becomes a trade-off: more performant models demand vast data, but their opacity exposes them to legal disputes. For companies choosing to retain control with on-premise infrastructure, the ability to document every phase of training and inference translates into a competitive and legal advantage. Beyond the legal outcomes, Midjourney's strategy raises an uncomfortable question for the entire industry: how aware are businesses of what ends up in their training data? Transparency, one way or another, is becoming the real battleground.