Principles Code (tentative) on the Protection of Intellectual Property and Transparency for Appropriate Use of Generative AI (Draft)

Applies To

Overview

The draft code was published by the Cabinet Office Intellectual Property Strategy Headquarters Secretariat as part of Japan's broader effort to apply its IP Strategy to the generative AI context. It is framed as soft law, meaning it does not carry the binding force of statute but is intended to shape industry conduct and may inform future regulation or enforcement posture. The code targets developers who build generative AI models and providers who deploy them to end users, setting expectations around copyright compliance, training data transparency, and output disclosure. Key provisions address how training data is sourced and documented, how rights holders are informed or compensated, and what disclosures must accompany AI-generated content. A public comment period ran from December 26, 2025 to January 26, 2026, after which the Secretariat was expected to review submissions before finalizing the code. Enterprises distributing or deploying generative AI products in Japan should treat compliance with this code as a reputational and pre-regulatory risk management obligation, particularly given Japan's active legislative environment around AI and IP.

Key Requirements

• •Developers and providers of generative AI must implement measures to identify and respect copyrighted works used in model training data.
• •Transparency obligations require disclosure of the general categories or sources of training data to enable rights-holder assessment.
• •AI-generated outputs must be disclosed as such in contexts where misidentification as human-created content could cause harm or confusion.
• •Providers must establish internal processes for receiving and addressing IP infringement claims related to both training data and model outputs.
• •Compliance with the code is expected on a voluntary basis pending finalization, but non-conformance may be referenced in future regulatory or judicial proceedings.
• •No formal penalties are specified under the draft code, as it is soft law; however, failure to comply may expose organizations to liability under existing Japanese copyright law.

What Your Organization Must Do

• →Audit all generative AI systems currently developed or deployed in Japan to assess training data sourcing practices against the draft code's transparency expectations.
• →Map training datasets to identify copyrighted third-party content and document the legal basis for its inclusion, such as licensing or statutory exceptions.
• →Implement output-labeling mechanisms that identify AI-generated content in customer-facing products or services distributed in Japan.
• →Establish or update an IP complaints intake process capable of handling rights-holder inquiries related to both training inputs and generated outputs.
• →Assign a designated compliance owner to monitor finalization of the code following the January 2026 comment period and track any amendments before adoption.
• →Update contracts with generative AI vendors and model suppliers to require equivalent transparency and IP-protection practices, ensuring downstream compliance obligations flow through the supply chain.