Last week, a Japanese industry organization representing publishers, including Studio Ghibli, formally sent a letter to OpenAI, demanding that it stop using copyrighted content to train its AI models without authorization.
Studio Ghibli, which created animated classics like Spirited Away and My Neighbor Totoro, is seeing its works affected by OpenAI’s generative AI products. In March of this year, following the launch of ChatGPT’s image generation feature, an online craze for “Ghibli style” image creation rapidly took off, with users transforming their photos or pet pictures into the Ghibli aesthetic. Even OpenAI CEO Sam Altman changed his X profile picture to a “Ghibli-fied” photograph of himself.
Now, with the wider release of Sora, OpenAI’s video generation application, the Content Overseas Distribution Association (CODA) of Japan has formally requested that OpenAI be prohibited from using its members’ content for machine learning training without authorization.
This demand is not without foundation. OpenAI adopts a “use first, negotiate later” strategy when dealing with copyrighted content, allowing users to easily generate images and videos involving copyrighted characters or deceased celebrities. Such actions have already drawn complaints from companies like Nintendo and the estate of Martin Luther King Jr., who have all expressed concern over the forgery risks that Sora might encourage.
OpenAI now faces a choice: comply with the demands or potentially face legal action. However, current US law still lacks a clear definition of whether using copyrighted material for AI training is legal.
To date, there are extremely limited precedents that judges can cite. The current US Copyright Act, enacted in 1976, has not undergone substantial revisions targeting AI technology. However, in a recent case heard by Federal Judge William Alsup, although Anthropic was fined for using copyrighted books to train its AI, the training behavior itself was not ruled illegal.
The Japanese CODA, however, points out that similar actions could constitute infringement in Japan. In a statement, the association emphasized: “Taking Sora 2 as an example, if a specific copyrighted work is reproduced or generated in a highly similar manner, CODA believes that the act of copying during the machine learning process may constitute infringement. Under the Japanese copyright system, the use of copyrighted works generally requires prior authorization, and there is no liability exemption mechanism that allows one to circumvent infringement responsibility through post-use objection.”
Hayao Miyazaki, the creative mastermind behind Studio Ghibli, has yet to publicly comment on the recent surge of AI-generated images in the style of his works.
