AI training data fight escalates as Studio Ghibli and Japanese publishers challenge OpenAI

Studio Ghibli and a coalition of major Japanese publishers, represented by the Content Overseas Distribution Association, demand that OpenAI stop training its models on their works. They argue that generative tools are learning and replicating protected styles without consent, and Japanese officials signal they are ready to enforce local copyright rules if complaints go unanswered.

Japan’s creative heavyweights are drawing a clear boundary. Studio Ghibli, known for films like Spirited Away and My Neighbor Totoro, has joined publishers including Bandai Namco and Square Enix to press OpenAI to halt the use of their content in model training. Their message, delivered through CODA in a formal letter, puts fresh global pressure on how tech companies source data for AI.

Why are Japanese creators pushing back now?

The trigger is simple to explain and hard to ignore. Popular generative systems can produce images and video that feel unmistakably “anime” or even “Ghibli-like.” That might delight fans. For rights holders, it looks a lot like unlicensed style replication at industrial scale. Millions of users are testing prompts that reference famous studios and characters, which blurs the line between homage and unauthorized copying. You can see how filmmakers are growing increasingly concerned as these results spread.

In its October 28 statement, CODA said tools such as the video generator often called Sora 2 can output visuals that closely resemble Japanese anime, games and manga. Their view is that such fidelity is unlikely without extensive exposure to those works during training. Studio Ghibli’s leadership has echoed the point, saying creators expect global tech firms to respect their rights, as reported by TechCrunch.

What is OpenAI’s position and how does the law apply?

OpenAI has discussed opt-out frameworks with some rights holders in the past. CODA argues that opt-out is not enough under Japanese law, which generally expects prior permission when copyrighted works are used beyond limited exceptions. This is a key difference from the United States, where fair use is broader and still being tested in court for AI training.

Japan’s legal backdrop is nuanced. The Copyright Act includes a data analysis exception that has been interpreted to allow text and data mining, but policymakers have been reassessing how far that should extend to generative AI. Government bodies have issued guidance for AI businesses and signaled that enforcement will follow if creators’ rights are ignored. Officials have publicly called anime and manga cultural treasures and warned that ongoing complaints will be taken seriously, as noted by Gamespot and other outlets.

What does this clash mean for AI and copyright worldwide?

This is not a niche fight. It is a test of how the world treats the data that powers AI. Courts in multiple countries are still wrestling with whether training on copyrighted works without a license can be fair, transformative use, or whether it crosses into infringement. High profile cases in the United States, including actions from media companies and artists, highlight how unsettled the rules remain.

Legal scholars in Japan and abroad note that the question is bigger than images. Voice cloning, music generation and text models all raise similar issues. As Professor Hiroshi Hasegawa of Keio University has argued, the AI era is forcing a fresh look at how we define copying and derivation for machine learning, a point reflected in coverage by Storyboard18.

Recommended tech

AI features are quickly moving into everyday laptops. The TechBull recommends the latest Copilot+ PCs, like the Lenovo IdeaPad Slim 3X AI Laptop, which bring on-device AI acceleration to creative and productivity workflows.

What happens next for OpenAI and Japan’s creative sector?

CODA wants two things. First, a stop to training on members’ works without permission. Second, a good faith process to address claims about model outputs that mirror protected styles or characters. With ministries indicating they are monitoring the situation, the stakes are clear. If the parties cannot find a compromise, lawsuits, regulatory action or both could reshape how datasets are sourced and how generative outputs are controlled.

For now, the standoff holds. Japanese creators have drawn a line that is as much about cultural stewardship as it is about economics in a world where AI is changing our digital world. However this dispute resolves, it will influence the playbook for AI training and licensing far beyond Japan.

FAQs