Japan Advances AI Transparency with Proposed Training Data Disclosure Code

Japan's government panel on intellectual property rights in the AI era approved a draft principle code on Tuesday that encourages generative artificial intelligence companies to disclose their training data and data collection methods publicly. The code is designed to protect intellectual property rights amid concerns over unauthorized use of copyrighted texts and images in AI training.

The draft code was presented during an online meeting of experts focused on intellectual property in the context of AI technologies. It was developed following the enactment of a law on AI-related technology in May 2025, which seeks to strike a balance between protecting rights holders and fostering technological innovation.

Under the proposed framework, the government will adopt a "comply or explain" approach. This nonbinding code allows generative AI businesses—including system developers and service providers—to either comply with the code or publicly explain their reasons for noncompliance. Companies that comply will be required to announce their adherence on their websites and notify the government accordingly.

The scope of the code extends to foreign businesses offering AI services within Japan, ensuring broad applicability across the industry.

The draft code is structured around three core principles. The first principle mandates that businesses disclose the generative AI models they employ, along with the training data and methods used to collect such data, on their websites for public access. However, disclosure of sensitive information is exempt from this requirement.

The second principle addresses copyright concerns by requiring businesses to disclose whether specific webpages are included in their training datasets upon request from copyright holders or other rights holders alleging infringement.

The third principle obligates businesses to respond to information requests from system and service users who have concerns about potential copyright infringement.

This initiative reflects Japan's proactive stance in addressing the complex intellectual property challenges posed by the rapid growth of generative AI technologies. By promoting transparency in AI training data, the government aims to mitigate unauthorized use of protected content while supporting continued innovation in the AI sector.

The principle code remains nonbinding but represents a significant step toward establishing industry standards for AI transparency and rights protection in Japan.