The Content Overseas Distribution Association (CODA), which advocates against piracy and represents Japanese intellectual property (IP) holders, including Studio Ghibli and Bandai Namco, issued a letter last week directed at OpenAI. The letter requests that OpenAI cease the use of its members’ content in the training of Sora 2, an AI model that reportedly generates content featuring copyrighted characters. CODA expresses concern that the replication process involved in machine learning could potentially lead to copyright infringement.
Since its launch on September 30, Sora 2 has produced a significant amount of content that references Japanese IP. This surge prompted Japan’s government to formally request that OpenAI refrain from replicating Japanese artwork. OpenAI’s applications have previously drawn on Japanese media; for instance, the launch of GPT-4o highlighted the generation of “Ghibli-style” images. Additionally, OpenAI CEO Sam Altman has a profile picture reminiscent of Studio Ghibli’s artistic style on social media.
Altman announced last month that OpenAI would adjust Sora’s opt-out policy for IP holders. However, CODA argues that the existing opt-out framework might conflict with Japanese copyright laws, as the laws generally require prior permission for the use of copyrighted materials. CODA elaborates that there is no provision in Japan’s copyright system that allows a party to evade liability for infringement through subsequent objections.
Now, CODA is asking OpenAI to “respond sincerely” to the copyright claims of its members and to discontinue the use of their content for machine learning without explicit authorization, which encompasses both Sora’s outputs and the use of Japanese IP as training data.
Source: https://www.theverge.com/news/812545/coda-studio-ghibli-sora-2-copyright-infringement