OpenAI’s Next Big Problem Isn’t Elon Musk — It’s Studio Ghibli

While everyone was rubbernecking the Musk vs Altman drama, OpenAI picked up a more immediate headache from Japan. After months of people pumping out Studio Ghibli-style AI art, Japan's Content Overseas Distribution Association (CODA) has stepped in and told OpenAI to cut it out: stop using its members' copyrighted works to train AI without permission, or expect legal trouble.

The Ghibli-style boom that OpenAI embraced

When ChatGPT added image generation, social media flooded with Ghibli-fied everything. Sam Altman even swapped in a Ghibli-style profile pic at one point. OpenAI leaned into the trend; engagement went up. One tiny problem: Studio Ghibli has never been on board with this. Co-founder Hayao Miyazaki has been open about his dislike of AI, and Ghibli's tastes run old-school and hand-drawn, not prompt-and-go.

CODA, which represents Japanese publishers and rights holders (including Studio Ghibli), drew a line: the group sent a formal demand letter telling OpenAI to stop training on members' content without explicit permission.

What CODA told OpenAI

The letter demands an immediate halt to training on copyrighted works owned by CODA members unless OpenAI gets permission first. And if OpenAI keeps doing it, CODA says legal action is on the table.

"Under Japan's copyright system, prior permission is generally required for the use of copyrighted works, and there is no system allowing one to avoid liability for infringement through subsequent objections."

Why this is a bigger risk in Japan than in the U.S.

OpenAI has built a scrape-first, ask-later reputation, which is why it is already getting sued by multiple outlets. In the U.S., the legal framework is still playing catch-up; the core Copyright Act dates to 1976, and courts are only now weighing whether training AI on copyrighted works is infringement. That uncertainty has not stopped lawsuits: The New York Times, a group of Canadian news publishers, and newspapers owned by Alden Global Capital have all filed cases challenging OpenAI's practices.

Japan is a different environment. There, using copyrighted works to train AI without prior permission can be treated as infringement. And unlike the U.S., you cannot just say "we will opt you out later" and expect that to shield you.

The timing: Sora enters the chat

CODA's letter lands just as OpenAI starts rolling out Sora, its text-to-video tool that can spit out animation-like clips in seconds, including shots that look very Ghibli-adjacent. Without proper licensing, that is basically asking for a courtroom date.

• CODA represents Japanese rightsholders (including Studio Ghibli) and has formally demanded OpenAI stop training on members' works without permission.
• OpenAI is already facing lawsuits from The New York Times, Canadian news publishers, and Alden Global Capital's newspapers over similar issues.
• Japan's system is stricter: permission first, or it is likely infringement; there is no get-out-of-jail-later mechanism.

Short version: The viral Ghibli trend was great for clicks, not so great for legal standing. If OpenAI wants to keep flirting with iconic styles, it is going to need actual agreements, not vibes.