Midjourney called worst AI model for Hollywood IP violations
A VN.ai report shared with Deadline says Midjourney v8.1 led 15 AI models in reproducing copyrighted characters and risky brand contexts.
Deadline reports that VN.ai, a startup that stress tests generative AI tools for copyright infringement, ranked Midjourney v8.1 as the worst of 15 popular AI models in its first quarterly report on studio IP. According to the report shared with Deadline, Midjourney reproduced copyrighted characters in 37% of text prompts and led rivals on “brand safety violations,” where characters appear in harmful contexts. Deadline says every tested model generated at least one copyright-protected character, even without naming one in the prompt. Midjourney did not respond; its copyright fight with Disney and NBCUniversal is ongoing, and the company argues training on copyrighted images is fair use.
Why it matters: For our creator/tech crowd, the key signal is that copyright testing is moving into model-by-model scorecards, not just lawsuits. If your workflow touches AI visuals, studio IP, or brand-safe content, this fight is getting practical fast.
Sources
This post summarizes reporting by the outlets above; read them for the full story. Fact-checked against the sources before publishing. Corrections: ashik@ashik.ai
Follow this story. We’ll show an alert on Ashik.FM when it is updated.