Background and Overview of the News
A controversy has arisen over the use of the generative AI “Sora 2”, developed by a U.S. tech company, to create and post videos featuring Japanese anime and other characters on the internet.
In response, 19 companies and organizations, including major Japanese publishers, issued a joint statement declaring that they “will not tolerate copyright infringement by generative AI.”
The statement explicitly demands that AI developers should:
- Obtain necessary permissions from rights holders,
- Ensure transparency of training data, and
- Provide fair compensation to rights holders when usage is permitted.
It also criticizes the mechanism requiring rights holders to opt out (“informing the developer if they do not wish to be included”), arguing that such a system itself could lead to copyright violations.
Meanwhile, OpenAI’s CEO Sam Altman announced plans to introduce a function in Sora 2 that allows users to specify how characters and other assets are used in more detail.
Why Japanese Anime Characters Have Become the Focal Point
Several factors explain this focus:
- Japanese anime and manga are recognized worldwide as cultural brands, carrying immense value as “cultural assets.” The Japanese government itself has described anime and manga as “irreplaceable national treasures.”
- Reports indicate that videos generated by Sora 2 have explicitly featured recognizable Japanese anime and game characters (e.g., Nintendo or Pokémon properties).
- Japan’s copyright system generally requires prior authorization, meaning AI developers can be criticized for “using or training without permission.”
- The quick response by rights holders—including publishers and the Japan Cartoonists Association—led to an early, coordinated institutional stance via a joint statement.
Thus, this issue is not merely about “AI’s technical ability to create,” but rather exposes deeper tensions among Japan’s content industry, cultural values, and legal frameworks.
The Core Issue: The Gap Between Generative AI and Copyright Law
This case highlights several interrelated problems:
- Relationship between training data, generated output, and copyright
Generative AI learns from vast existing datasets (images, videos, text, etc.). The key question is whether training or output generation using copyrighted works (e.g., characters, anime, games) constitutes copyright infringement.
Japanese law defines “reproduction,” “adaptation,” and “public transmission” as infringing acts, but it remains unclear whether AI learning and generation fit into these categories.
- Opt-out (post-use exclusion) vs. Opt-in (prior authorization)
Some AI companies, including OpenAI, initially adopted an opt-out model, under which rights holders must explicitly request exclusion. Rights holders argue this framework is flawed and advocate for a prior authorization (opt-in) system.
- Compensation and transparency challenges
If AI can generate works resembling existing characters or styles, rights holders risk losing legitimate revenue opportunities. Lack of transparency about the origin of training data further exacerbates this issue.
- Anime and manga as global cultural assets
Japan’s anime and manga industries are deeply intertwined with global licensing, brand management, and fan culture. Generative AI’s ability to mass-produce “hybrid” videos combining existing characters with new contexts could dilute brand value and cause rights confusion.
Significance of the Joint Statement by Publishers and Organizations
The 19 publishers and industry groups emphasized:
- A reaffirmation that copyright infringement will not be tolerated.
- AI developers must obtain permissions from rights holders.
- Transparency in training data usage.
- Fair compensation for authorized uses.
- A rejection of the opt-out framework as potentially infringing.
The statement’s significance lies in:
- Presenting a unified position among rights holders regarding AI.
- Moving beyond mere opposition to propose constructive demands such as “transparency” and “fair compensation.”
- Signaling Japan’s content industry’s readiness to establish governance frameworks for the AI era.
In essence, the statement shifts the principle from “exclusion after the fact” to “authorization before use,” suggesting a new contractual and institutional paradigm for the AI age.
The Response and Limitations of AI Developers (OpenAI)
OpenAI’s situation can be summarized as follows:
- Shortly after Sora 2 was released, videos featuring imitations of Japanese anime and game characters appeared online.
- OpenAI initially adopted an opt-out approach, allowing usage unless rights holders explicitly objected.
- Following backlash from Japan and abroad—including requests from the Japanese government to protect anime and manga as “irreplaceable national assets”—CEO Sam Altman announced plans for more granular control features regarding character use.
However, major limitations remain:
- There are no formal legal obligations for developers to obtain prior authorization or verify data rights clearance.
- The origins and legal status of training data remain opaque.
- “Fine-grained control features” may not realistically prevent the mass generation and rapid online spread of infringing content.
- Power imbalances persist between content creators and large AI platforms.
Thus, while initial steps have been taken, the convergence of technology, law, and business norms around generative AI and copyright is still unresolved.
Future Outlook and Perspectives for Creators and Businesses
- For creators and rights holders:
Review how your works might be used by AI. Establish clear internal policies on granting or withholding permissions.
When granting licenses, define the scope of use, compensation mechanisms, and distribution control in contractual terms.
This era invites a rethinking of sustainable business models that protect brand value and creator-fan relationships.
- For AI developers and platform operators:
Establish transparent permission processes and ensure that only cleared materials are used for training and generation.
Disclose the origins, types, and quantities of training data to enhance accountability.
Develop technical and contractual filters to prevent outputs that resemble or misuse existing IP.
Explore models that enable rights holders to share in AI-driven creative value.
- For policymakers and institutions:
Copyright frameworks—built for “human-created works”—are lagging behind AI’s data-driven paradigm.
Future policy may introduce new rules on training data authorization, output similarity standards, and compensation schemes.
Given Japan’s focus on protecting cultural assets like anime and manga, it aims to take a global lead in AI-era IP governance.
Early adaptation to these shifts will be crucial for maintaining competitiveness in both creative and AI industries.
Toward Coexistence Between Cultural Heritage and Digital Innovation
This controversy is not just about “AI using characters without permission.” It raises fundamental questions:
- How should technology treat creative works built through cultural effort and artistic legacy?
- How can society embrace innovation without sidelining existing copyright systems and creator economies?
- How do we balance the democratization of creativity enabled by AI with fair compensation and respect for creators?
The pursuit of “technological innovation × cultural heritage protection” will be a defining challenge for Japan’s content and AI industries alike. Each of us—creators, businesses, and users—must reconsider where we stand, what we protect, and how we collaborate in this evolving landscape.