A prominent Japanese trade consortium, representing some of the most influential creative enterprises in the country—including globally recognized studios such as Studio Ghibli, Square Enix, and Bandai—has recently issued a formal communication to OpenAI. The letter, dated October 28, expresses grave concerns regarding possible instances of copyright infringement associated with OpenAI’s activities. This collective of renowned content producers, united under the Japan-based organization known as CODA (the Content Overseas Distribution Association), has positioned its correspondence as a measured but unmistakable signal to one of the world’s leading developers of artificial intelligence technology that the boundaries between original Japanese intellectual property and algorithmically generated output may be in jeopardy.
The document, according to CODA’s announcement, contains several key observations focusing on the growing resemblance between the visuals generated by OpenAI’s Sora 2 text-to-video model and well-known forms of Japanese visual and narrative expression. CODA argues that certain Sora 2 videos bear a notable likeness to what it describes broadly as “Japanese content,” thereby raising concerns that such outcomes might not be purely coincidental. Within the same letter, the organization articulates two distinct requests. First, it asks that OpenAI abstain from utilizing any copyrighted materials belonging to CODA’s member companies as training data unless explicit, prior authorization is obtained. Second, it urges that OpenAI offer a “sincere response” whenever an individual member of CODA submits a complaint or raises a grievance regarding potential copyright violations. It is worth emphasizing that this communication stops short of aggressive or adversarial posturing—there are no explicit demands for immediate action, nor does the letter contain language that could be construed as a direct legal threat. Instead, CODA’s tone remains formal, deliberate, and conciliatory, seemingly intended to encourage dialogue rather than confrontation.
Sora 2 itself, OpenAI’s state-of-the-art text-to-video model, had been officially introduced toward the end of September. Its unveiling was met with reactions that ranged from astonishment at its technical sophistication to alarm over the apparent disregard for creative originality. Almost instantaneously, the public sphere—particularly artists, legal experts, and fans of Japanese media—was swept into what has been described as an onslaught of “copyright chaos.” Among the most controversial examples were visual outputs unmistakably evocative of major Japanese franchises and artistic styles: familiar creatures resembling those from the Pokémon universe, imagery reflecting the dystopian cinematic atmosphere found in the works of game designer Hideo Kojima, and sequences uncannily suggestive of animation derived from Studio Ghibli’s distinctive aesthetic. Such resemblances exacerbated anxieties about how closely generative AI systems may be imitating, or even reproducing, copyrighted patterns from their training sets.
What stands out in CODA’s communication is the subtle difference in legal reasoning and rhetorical tone compared to the complaints typically voiced in the United States when similar concerns arise. The Japanese association frames its accusation in a manner deeply anchored in its domestic legal culture. According to the organization’s statement, the similarity between Sora 2’s output and established Japanese creations is likely “the result of using Japanese content as machine learning data.” Furthermore, CODA asserts that when copyrighted materials are incorporated into machine learning processes in a way that allows them to be partially reconstructed or recognizable in the resulting output, “the act of replication during machine learning may itself constitute copyright infringement.” This stance reflects a distinctive and evolving interpretation of what it means to copy or reuse creative works in the age of artificial intelligence.
Japan’s Copyright Act includes a provision that may clarify CODA’s position—Article 30-4, a section specifically addressing the use of creative works for technological purposes such as artificial intelligence and data analysis. The clause outlines that the “exploitation of a work for purposes other than enjoyment,” for example, during AI training or analytical processing, is, in principle, permissible without the copyright holder’s consent. This legal flexibility, detailed in various governmental fact sheets, helps explain why CODA has opted for a restrained and courteous initial approach rather than pursuing immediate litigation or public condemnation. Japan’s statutory framework allows for certain noncommercial or research-oriented uses of copyrighted material, giving technology developers leeway to experiment within defined boundaries.
Nonetheless, CODA underscores that the basic principle within Japan’s legal tradition still favors strong protection of intellectual property rights. In its statement, the organization emphasizes that “prior permission is generally required for the use of copyrighted works,” reiterating that Japan does not operate a system under which alleged infringers can simply rectify their actions retrospectively or evade liability through later negotiations. In other words, the association contends that the moral and legal responsibility for obtaining consent should occur before any material is used for machine learning, not after a dispute emerges. CODA’s carefully worded letter thus signals an appeal for mutual respect between technological innovation and artistic integrity, balancing a recognition of AI’s transformative capabilities with a strong defense of creators’ rights and cultural identity.
Sourse: https://gizmodo.com/a-trade-group-that-includes-studio-ghibli-just-slapped-openai-with-a-letter-2000681040
