Midjourney asks to broaden AI-discovery in Hollywood copyright clash

Midjourney is pushing to force Disney, Universal, and Warner Bros. to disclose their internal AI practices, including training datasets and model weights, as part of its defense in a high-stakes copyright lawsuit. The outcome of this discovery dispute could set a precedent for how much internal AI usage is subject to disclosure in litigation, directly affecting AI builders who rely on fair use arguments or train on publicly available data.

What happened

In June, a magistrate judge limited Midjourney's discovery to consumer-facing AI applications only, ruling that broader information about the studios' internal AI use was irrelevant to infringement. Midjourney has now asked Judge John Kronstadt to overturn that decision. Midjourney's attorney argued that evidence of the studios' own AI training on unlicensed copyrighted data would support its fair use and unclean hands defenses and show industry custom. The studios have called the request a fishing expedition and maintain they only want Midjourney to stop copying their characters.

Why AI builders should care

This case tests whether companies that sue AI developers for copyright infringement can shield their own internal AI training practices from discovery. If Midjourney succeeds, it could force plaintiffs to open their own training data and model development processes, creating a more level playing field for fair use defenses. For AI builders, this could influence how they document internal AI usage and what they might need to disclose in future litigation.

Practical implications

A broader discovery standard would mean that any company bringing a copyright claim against an AI product could be required to reveal its own AI training datasets, model weights, and business plans. This could deter some lawsuits and encourage more transparency around training data provenance. It also raises questions about how studios and other content owners use AI internally while challenging external AI use.

Caveats

The case is ongoing and the ruling could be appealed further. The studios strongly oppose the motion, and the magistrate judge's initial order suggests courts may be reluctant to allow broad discovery into internal AI practices. The outcome may also depend on how Judge Kronstadt interprets relevance and proportionality under federal discovery rules.

FAQs

What is Midjourney asking Disney, Universal, and Warner Bros. to disclose about their AI use?

Midjourney is seeking AI business plans, research reports, training datasets, model weights, internal tool use data, and board meeting presentations about AI. The studios have only agreed to turn over information about consumer-facing AI applications.

How could internal AI training data and model weights affect the Midjourney copyright case?

Such materials could support Midjourney's fair use and unclean hands defenses by showing that the studios themselves train AI on unlicensed copyrighted data, potentially establishing an industry custom.

What is the significance of fair use and unclean hands defenses in this AI copyright battle?

Fair use argues that using copyrighted material for transformative purposes is not infringement. Unclean hands claims that the plaintiff is engaging in similar misconduct, barring them from seeking relief. Evidence of the studios' own AI training on unlicensed data could strengthen both defenses.

How might broadened discovery impact future AI copyright litigation?

If broader discovery is allowed, it could set a precedent for accessing internal AI development practices in similar lawsuits, influencing strategy and transparency expectations for both plaintiffs and defendants.