Disney, Universal and Warner Bros. Discovery sue MiniMax over Hailuo AI

Background

Hailuo AI is the video generation product of MiniMax, a Chinese artificial intelligence company. Its video-01 model launched in September 2024, part of a wave of systems that turn a written prompt into a short moving clip rather than a still image.^[1]

Video models of this kind learn by processing very large quantities of existing footage. From that material they pick up patterns — how a face is put together, how a cartoon character is drawn, how a broadcaster's logo is laid out — and can then produce new images that resemble what they have seen. That is also where the trouble starts: a prompt that names or describes a well-known character may come back with something close to the character itself.

Signs of that risk had already been reported. In December 2024, Broadcast reported that Hailuo AI could reproduce the logos of Channel 4, Channel 5 and ITV, three British broadcasters.^[1] A logo is a small, distinctive piece of graphic design, and a model that can render one on request has clearly absorbed branding it did not own.

The lawsuit

In September 2025, Disney, Universal and Warner Bros. Discovery filed a copyright lawsuit in the United States against MiniMax.^[1] The Standard reported that the suit was filed on a Tuesday and dated 16 September 2025, in a California district court, and that the plaintiffs were Walt Disney, Comcast's Universal and Warner Bros Discovery.^[2]

The studios alleged that Hailuo AI infringed their copyrighted characters.^[1] The complaint therefore centred on the output of the tool — what the model produced when users asked it for video — rather than on the software itself.

What the studios said about warnings

The studios also said that MiniMax had ignored their requests to put reasonable safeguards in place against infringement.^[2] That claim shifts part of the dispute from the question of what a model can do to the question of what its maker does about it. Filters on prompts, filters on outputs, and blocking requests for named characters are the usual ways a developer tries to lower the chance that a general-purpose generator is used to reproduce protected work. The studios' position, as reported, was that they had asked for measures of that sort and had not got them.

Why it matters

Copyright law in the United States gives the owner of a work exclusive rights over how it is copied and adapted, with limited exceptions. Generative video sits awkwardly against that framework: the model does not store a copy of any single film or drawing in a way a court can point to, yet it can produce imagery that reads as a particular studio's character. Courts have been asked to work out where learning from existing works ends and infringement of them begins.

The suit also put a Chinese company's consumer video product directly in front of a United States court, with three of the largest holders of screen characters as the opposing side. For users of tools like Hailuo AI, the practical stake is what prompts will be allowed and what output will be blocked; for developers, it is how much filtering and monitoring they are expected to build in, and whether being asked once is enough to make the failure to act a wrong in itself.

Afterwards

MiniMax went public in Hong Kong four months later, on 9 January 2026.^[1] The listing came after the suit had been filed, and it placed the company's shares in front of public investors while the copyright dispute was still working its way through a California court.