The Japanese Cabinet Office released a draft guideline for generative AI developers on August 18, requiring developers to disclose training data and collection methods. The draft aims to establish a balanced framework that regulates the reasonable use of intellectual property by developers, supports technological innovation, and protects creators' rights.

The draft itself is not legally binding, but it clearly urges AI developers to maintain transparency, publicly explaining how training data is collected, and suggests avoiding the use of piracy websites as sources.

The framework encourages disclosure of whether the training set includes copyrighted materials such as comics, music, and movies, and requires the establishment of an internal notification system—when training data includes copyrighted works similar to the generated content, users must be alerted. This soft constraint sends a signal: in the copyright-sensitive Japanese content industry, the space for AI companies to operate with "black-box training" is being continuously reduced.