Japan proposes disclosure requirements for AI training data

A Japanese government panel has approved a draft “principle code” for generative artificial intelligence businesses to protect intellectual property rights. The plan urges companies to disclose their AI training data and collection methods to the public to address concerns regarding unauthorized use of copyrighted texts and images.

The proposed framework utilizes a “comply or explain” approach. Under this nonbinding code, system developers and service providers—including foreign businesses operating in Japan—can choose to either comply with the principles or publicly explain their reasons for non-compliance.

The draft includes three core principles: requiring businesses to make training data and collection methods publicly accessible on their websites, disclosing specific webpage usage upon request from rights holders alleging infringement, and responding to user inquiries regarding copyright concerns.