Japan asks OpenAI to get approval for Sora 2 anime content

Japan’s government has asked OpenAI to seek prior approval from intellectual property owners before using their works in Sora 2, its short-form AI video app, amid concerns over copyright infringement.

Sora 2 allows users to create realistic AI-generated videos using text prompts, but officials and critics have raised alarms about insufficient protections for copyrighted characters, especially from anime and manga.

Digital minister Masaaki Taira said the government requested OpenAI shift from an opt-out to an opt-in system for copyright holders and introduce compensation for use of their characters.

Taira also said OpenAI has implemented a process for copyright holders to request content removal.

The app reached 1 million downloads in five days on the US App Store, but Nintendo said it would continue to act against intellectual property infringement, regardless of the technology used.

🔗 Source: The Japan Times

🧠 Food for thought

Implications, context, and why it matters.

Japan’s opt-in push is a policy ask, not law yet

Japan asked OpenAI to move from opt-out to opt-in (use of protected material only with prior permission) under the AI Promotion Act, a framework law that sets goals over penalties 1.
digital minister Taira said Japan could invoke the Act if OpenAI refuses, yet the Act functions as soft law that favors voluntary compliance over enforceable duties 2.
Japan’s copyright law permits non-expressive training without authorization. The carve-out may not cover fine-tuning such as Low-Rank Adaptation (LoRA) or training on databases that imitate specific styles with expressive intent 2.
The government’s layered approach uses nonbinding guidelines with existing copyright interpretation. That mix creates ambiguity for AI platforms, with enforcement tied to court rulings and administrative guidance 2. Public procurement counts. Environmental, Social and Governance (ESG) assessments matter too 2.

Anime IP tools

Over 90,000 AI anime images appeared on sharing sites, with about 2,500 matching multiple features from original works 3.
Character recognition APIs can serve generative platforms with models that flag traits such as facial structures, color palettes, or signature poses to ease the black box issue when training data stays undisclosed 3.
Licensing workflow platforms can link IP holders with AI companies as the government pushes a mechanism to compensate rights owners when their characters are used 4.
Products that bridge the gap between abstract character ideas and specific visual designs protected under Japanese law can lower infringement risk while keeping creative use possible 5.