Japan Adopts Nonbinding Principle Code to Push AI Training-Data Transparency

Overview

• The government formally adopted the Principle Code on Tuesday, Aug. 25, 2026, and will require firms that accept the code to notify authorities and post prescribed transparency information on their websites.
• The code asks AI companies to publish the model training process, types of training data and methods used to collect that data so outside parties can see how models were built.
• Under the guideline, firms must disclose to copyright holders or AI users whether training sets may contain material that could lead to copyright infringement when certain conditions are met.
• The measure has no legal force or penalties, which raises questions about whether voluntary transparency will be followed and how companies can protect trade secrets when they disclose details.
• The code explicitly covers overseas operators and could push foreign firms to reveal more about data practices, but its real impact will depend on voluntary uptake and any future regulations that set binding rules.