Japan’s draft “Principle Code” on intellectual property protection and transparency for generative AI (see our earlier post for a summary of the code) has attracted significant attention, with more than 2,000 consultation responses received from businesses, industry groups and rights holders. The responses suggest that transparency will play a central role in Japan’s approach to generative AI. At the same time, they highlight significant differences in views on how that transparency should be implemented in practice.
1. Broad support for transparency, but not for the same model
There is strong alignment on one point: transparency matters. It is widely seen as important for building trust in generative AI systems, particularly in relation to training data and outputs. It may also serve a practical function in the market, helping customers and counterparties assess AI providers. That alignment, however, breaks down at the level of implementation. The responses reveal clear differences in views on the appropriate scope, level of detail and who should bear disclosure obligations.
2. A divide between rights holders and industry
Two distinct perspectives emerge. Rights holders generally advocate for stronger and more meaningful transparency. Many suggest that high-level disclosures will not be sufficient and call for greater visibility into training data, improved traceability and, in some cases, mechanisms to assess whether specific works have been used. Some also point to interests beyond copyright, such as image rights and attribution.
Industry participants, by contrast, emphasize feasibility, confidentiality and competitiveness. More detailed disclosure requirements may risk exposing proprietary information or trade secrets. Respondents also highlight technical limitations, noting that it may not be possible to identify whether specific data was used in training or to trace outputs back to particular sources. The operational burden is another concern, particularly where disclosure obligations involve responding to broad or repeated inquiries.
3. Structural and policy questions
The responses also raise broader questions about how the framework should operate. One issue is its scope. Some respondents suggest that obligations should focus primarily on model developers, rather than applying across the entire value chain where downstream providers may not have access to the relevant information. Cross-border considerations were also raised. If the framework primarily affects domestic businesses, it may create competitive imbalances with overseas providers. In addition, although the Principle Code is formally non-binding, a number of responses note that it may have practical effects if it becomes relevant to procurement, funding or other policy incentives.
4. Interaction with Japan’s existing framework
Some submissions question how the draft aligns with Japan’s broader legal and policy landscape. Japanese copyright law already permits certain uses of copyrighted works for machine learning, and additional transparency requirements may be seen as introducing new obligations. Some respondents also observed that the draft may need to be considered alongside Japan’s traditionally flexible and internationally-aligned approach to AI governance.
5. Inquiry-based mechanisms
One of the more debated features is the proposal to require responses to inquiries from rights holders and users. Supporters view this as a way to reduce gaps in information between AI providers and rights holders and provide a practical mechanism to assess potential infringement. Others question its effectiveness, noting that existing legal processes already provide avenues for disclosure and that such mechanisms may be difficult to operate in practice.
6. What this means for business
For businesses developing or deploying generative AI in Japan, the consultation responses provide a useful indication of how stakeholder expectations may evolve. Transparency around training data and governance is likely to remain a focus, although the scope and level of disclosure remain under discussion. The responses also suggest that companies may need to consider how they balance transparency with the protection of confidential information, as well as how they approach potential inquiries from rights holders and users.
7. Takeaways
The consultation responses show both broad support for transparency and significant divergence on how it should be implemented in practice. While the final form of the Principle Code remains uncertain, it is clear that key questions — around the scope of disclosure, protection of confidential information and the operation of inquiry-based mechanisms — will need to be resolved. For now, the responses are a useful indicator of the issues under debate and the range of stakeholder expectations that may shape the final framework.
