The Japanese organization CODA, representing rights holders such as Studio Ghibli and Bandai Namco, has approached OpenAI demanding to cease using its members’ content for training the AI Sora 2. CODA stated that the “opt-out” policy for model training might violate Japanese copyright law. Japanese legislation typically requires prior consent from copyright holders for the use of their works and does not allow avoiding liability for violations through subsequent objections.
The organization emphasized that during the launch of Sora 2 on September 30, the AI generated a large amount of content using Japanese intellectual property. This led to an official request from the Japanese government to OpenAI to stop copying Japanese artistic works. CODA also noted that similar incidents have occurred with other OpenAI products, for example, during the launch of GPT-4o, many images in the style of Studio Ghibli appeared.
In October, Sam Altman announced that OpenAI would change the “opt-out” policy for intellectual property owners regarding Sora. However, CODA believes that the mere possibility to opt-out does not meet the requirements of Japanese law, as permission must be obtained before using the content.
CODA is now demanding that OpenAI cease using its members’ content for machine learning without prior consent. The demand concerns not only the results of Sora 2’s work but also the use of Japanese intellectual property as training data for AI.
