Japan’s Entertainment Titans Allege OpenAI’s Sora 2 Trained on Their IP Without Permission

Japanese entertainment heavyweights have formally challenged OpenAI over Sora 2, the company’s AI video generator, demanding an immediate halt to the use of copyrighted works in its training data. The warning comes from the Content Overseas Distribution Association (CODA), an anti-piracy group established in 2002 to protect Japanese media internationally. CODA’s membership includes major industry names such as Square Enix, Bandai Namco, Kadokawa (the parent company of FromSoftware), Studio Ghibli, and Cygames.

Sora 2 was announced on September 30 and rolled out via an invite-only iOS launch on October 1. In its letter, CODA asserts that a significant number of Sora 2’s outputs appear to closely mirror Japanese media and imagery, leading the association to conclude that Japanese works were likely used as training data without authorization. CODA also argues that when specific works are recreated or closely imitated by an AI model, the act of copying during the machine learning process itself may constitute copyright infringement under Japanese law.

A key point of contention is OpenAI’s reported opt-out mechanism for rights holders. CODA maintains this approach is incompatible with Japan’s legal framework, where prior permission is generally required to use copyrighted material. In other words, creators shouldn’t have to chase down platforms after the fact to protect their intellectual property.

Public concern intensified after a viral clip showed a video of Sam Altman interacting with Pokémon characters, captioned, “I hope Nintendo doesn’t sue us.” Although Nintendo is not part of CODA, the moment underscored how easily beloved Japanese franchises could be invoked by generative tools, driving fresh scrutiny from rights holders.

CODA’s demands are straightforward. The association wants OpenAI to ensure that member companies’ works are not scraped or used as training data without express consent. It also expects OpenAI to respond promptly and transparently to any copyright-related complaints involving Sora 2’s outputs. Failure to do so could open the door to legal challenges, particularly in a jurisdiction that puts strong emphasis on prior authorization for use of creative content.

This dispute unfolds as Japan’s game and entertainment sectors increasingly explore AI. A recent survey cited by local reports indicates that more than half of the companies polled—among them Capcom and Level-5—are experimenting with artificial intelligence for development workflows. The push-pull is clear: studios are investigating AI’s potential to accelerate production, while simultaneously seeking firm guardrails to prevent misappropriation of their IP.

The outcome of CODA’s challenge could have far-reaching implications. If Japanese authorities or courts embrace the view that training on copyrighted material without permission is infringing, it may reshape how AI video models are built, licensed, and deployed in one of the world’s most influential media markets. For now, the message from Japan’s entertainment giants is unambiguous: innovation should not come at the expense of creators’ rights.