Ghibli Challenges OpenAI's AI Training

What happened

CODA, representing Studio Ghibli and other Japanese publishers, demanded OpenAI cease using their copyrighted content to train AI models, specifically Sora 2. This action targets a 2018 Japanese copyright law revision, which inadvertently permits foreign firms like OpenAI to incorporate anime, games, and cultural works into AI training datasets without explicit permission. The demand follows concerns over ChatGPT's image-generation mimicking Ghibli's style. CODA seeks governmental intervention and OpenAI's response to infringement claims regarding Sora 2 outputs.

Why it matters

The identified legal loophole in Japan's 2018 copyright revisions introduces a significant operational constraint by reducing explicit control over intellectual property used in AI model training. This creates a visibility gap for content owners regarding the inclusion of their works in datasets, increasing exposure to potential copyright infringement. Legal and intellectual property management teams face heightened due diligence requirements to monitor and address unauthorised content usage by AI developers like OpenAI.