The image generator getting sued for IP theft just asked its accusers to open their own AI books.
The Summary
- Midjourney filed discovery motions demanding Warner Bros., Paramount, and Universal disclose their internal AI usage as part of their copyright infringement lawsuit against the company
- The legal strategy flips the script: instead of defending what Midjourney trained on, force studios to explain what THEY'RE building with AI
- This could expose the hypocrisy of studios suing AI companies while quietly deploying the same tech to replace workers
The Signal
Midjourney isn't playing defense anymore. In discovery requests filed this week, the company wants Warner Bros., Paramount, and Universal to detail every AI tool, model, and dataset they use internally. What did they train their models on? Which scripts, storyboards, or footage? Who gave permission?
The three studios sued Midjourney in 2025, claiming the image generator infringed their copyrights by training on proprietary imagery without permission. Standard Hollywood lawsuit. Except Midjourney's countermove exposes an uncomfortable truth: studios have been experimenting with generative AI for concept art, pre-visualization, and script analysis for years.
"If training on copyrighted material is theft, then every studio with an internal LLM has some explaining to do."
The timing matters. Hollywood is still raw from the 2023 strikes, where AI guardrails became a core demand. Writers and actors negotiated contract language limiting AI's role in creative work. But those contracts say nothing about what studios can build behind closed doors for "research purposes" or "internal tools."
Midjourney's discovery request targets three areas:
- What AI models the studios have licensed or built in-house
- What data those models were trained on, and whether rights were cleared
- How studios are using AI in active production workflows today
If the court grants the motion, we'll get a rare look at Hollywood's private AI operations. The studios have been loudly anti-AI in public while reportedly testing tools to streamline VFX, automate rotoscoping, and generate background elements. That's not hypocrisy, exactly. It's the difference between outsourcing creative decisions to a third party and keeping the tech in-house under NDA.
But Midjourney is betting a judge won't see it that way. If Warner Bros. trained a model on its own back catalog to generate concept art, did it get permission from every director, DP, and production designer whose work fed that model? Probably not. Employment contracts don't typically grant studios the right to use your creative output as training data for machines that might replace you.
The Implication
This case could force Hollywood to admit what everyone already knows: studios aren't fighting AI, they're fighting AI they don't control. If discovery proceeds, expect leaked memos about internal image generators, voice cloning pilots, and script analysis tools that never made it into press releases.
For builders in the agent economy, watch how studios respond. If they settle quickly, it signals they have something to hide. If they fight discovery, it means they think their internal AI usage is defensible under different rules than what Midjourney does. Either way, we're about to learn whether "do as I say, not as I do" holds up in court.