Midjourney named worst offender in report on AI copying Hollywood characters
Midjourney v8.1 copies Hollywood characters more readily than 14 rival AI models, according to the first quarterly report from VN.ai on studio intellectual property, as reported by Deadline.
The report, released on 14 October, named Midjourney v8.1 the worst offender for reproducing images of copyrighted characters. It also ranked the model highest for brand safety violations, a term the industry uses for behaviour that places familiar characters in harmful contexts such as violence or drug taking.
The numbers were stark. Midjourney produced an image that broke brand safety rules in 37% of text prompts, or about once in every three tries, according to VN.ai's testing cited by Deadline. The example was blunt. One prompt returned The Joker smoking cannabis while wearing a dress resembling Elsa's wardrobe from Frozen.
The problem was not limited to one company. Every model tested generated at least one copyright-protected character even when the character's name was not mentioned in the prompt, the report found. That included Google's Nano Banana and ByteDance's Seedance 2.5, which each produced a protected character without a direct name reference, according to Deadline.
ByteDance's Seedance 2.5 improved significantly after receiving cease-and-desist letters from Hollywood studios, the report said. A cease-and-desist letter is a formal legal warning demanding that an activity stop. For readers, this means filters can change quickly when lawyers get involved.
Video brought its own risks. MiniMax H3 generated videos that violated studio brand safety standards 55% of the time when an image was supplied as part of the prompt, according to the VN.ai findings. That is more than half of those image-led tests.
What makes this stand out is the courtroom context. Disney and NBCUniversal have accused Midjourney in an ongoing copyright lawsuit of being a "bottomless pit of plagiarism," as reported by Deadline. Midjourney has argued that training its models on copyrighted images is permitted under fair use, a US legal rule that can allow limited use of protected work without permission.
The lawsuit, first filed on 11 June according to earlier reporting from Georgetown Law's Tech Institute, remains active. Midjourney is also appealing a magistrate judge's mid-June 2026 order that limited its ability to demand details about how the studios themselves use artificial intelligence, according to Variety. The judge restricted that disclosure to consumer-facing AI.
VN.ai was founded by Nelson Chu, according to Deadline. The company describes its purpose as controlling how AI uses intellectual property, and says it tests whether models can reproduce a character without being given the name.
