Japanese content creators have expressed strong dissatisfaction and taken action against the behavior of AI giant OpenAI, which has used their copyrighted materials to train generative AI models without permission.

Last week, a Japanese industry organization representing multiple publishers, including the renowned animation studio Studio Ghibli, called CODA (Japan Association for Overseas Content Distribution), sent a letter to OpenAI, formally requesting it to stop using its members' content for machine learning without authorization.

OpenAI

 Studio Ghibli is the "hard hit," CEO also used AI-generated images

Studio Ghibli (known for works such as "Spirited Away" and "My Neighbor Totoro") has been particularly affected by OpenAI's products. In March this year, after the release of OpenAI's image generator, "Studio Ghibli-style" selfies and pet photos quickly became a trend. Even OpenAI's CEO Sam Altman once changed his profile picture on X to an image in the "Studio Ghibli style."

Now, with more users able to use OpenAI's video generator Sora, CODA has requested that OpenAI not use its members' content for machine learning.

⚖️ "Cutting off first, asking forgiveness later" model sparks public anger, deepfakes are easy to create

CODA's request is not baseless. OpenAI has adopted a **"cutting off first, asking forgiveness later"** approach when dealing with copyrighted content. This practice has not only led to complaints from institutions such as Nintendo but also caused dissatisfaction from the Martin Luther King Jr. Estate, as users can easily generate copyrighted characters or even create deepfakes of deceased celebrities.

CODA emphasized that this approach could be considered copyright infringement in Japan. According to Japan's copyright system, using a copyrighted work usually requires previous permission, and "there is no system that allows avoiding liability for infringement by raising objections afterward." CODA further pointed out that in cases like Sora, "when a specific copyrighted work is copied or generated in a similar manner as output, CODA believes the copying during the machine learning process may constitute copyright infringement."

💥 Unclear copyright laws, Hayao Miyazaki once expressed "disgust"

Although OpenAI can decide whether to comply with the request, if it does not, the affected parties may take legal action. However, U.S. law still lacks clarity on using copyrighted materials for AI training, with copyright laws unchanged since 1976, leaving few precedents. A recent ruling by a U.S. federal judge determined that Anthropic company did not violate the law by using copyrighted books to train AI, despite being fined for using pirated books for training.

Hayao Miyazaki