Hello OSI community,
I am writing from Japan to flag an ongoing public consultation by the Cabinet Office of Japan. The Cabinet Office has published a draft “Principle-Code for Protection of Intellectual Property and Transparency for the Appropriate Use of Generative AI” (provisional title), and is currently accepting public comments.
Call to action (deadline): Public comments are open until January 26, 2026, 23:59 Japan time (JST). Submissions are accepted in English as well as Japanese.
Why this matters outside Japan
The draft explicitly says it applies even if a business does not have its head office in Japan, as long as the generative AI system or service is provided to Japan (including when it is available to Japanese nationals).
This is why Open Source developers and global service providers should pay attention.
What in the draft could affect Open Source and global providers
1) Very broad scope of “developer” and “provider”.
A “generative AI developer” is defined broadly as someone responsible for building a “generative AI system”, including a model and the system infrastructure of the model, plus related components, and who provides all or part of that system to the public.
Depending on interpretation, this could create uncertainty not only for model publishers, but also for projects and platforms in the wider ecosystem.
2) “Comply or explain”, but with publication and annual updates.
Businesses that accept the principles are expected to publish an acceptance statement and the implementation status for each principle on their corporate website, submit the same to the Cabinet Office in a prescribed format, and update it annually. The Cabinet Office plans to publish a list of submitters and links.
The draft also anticipates government incentives based on disclosures and initiatives, which can turn a voluntary code into a practical requirement over time.
3) Deep transparency expectations (including crawling practices).
The accompanying “specific examples” document includes disclosures such as crawler purpose, collection period, crawler identifiers, and whether third-party crawlers are used.
It also includes respecting paywalls and robots.txt, publishing measures with user agents, and notifying changes.
It further references retention of learning logs and traceability practices as examples.
4) Disclosure request frameworks (Principle 2 and Principle 3).
Under Principle 2, the draft contemplates responding to requests that include disclosure of certain “control information” (for example, whether training/validation data includes URLs queried by the AI business). It also states that even if requested matters are considered trade secrets, businesses are expected to first seriously consider and discuss the matter.
Principle 3 similarly contemplates requests by users, with the prompt and generated output provided, and requests related to URLs that host allegedly similar content.
5) The “Open Source” exception may not reduce the burden in practice.
The draft provides an exception where disclosure is difficult due to the use of Open Source software, allowing a form of substitute disclosure by clarifying the use of Open Source software and license details.
However, this may not meaningfully reduce uncertainty for Open Source AI projects. In particular, projects that already publish model weights, training data, or other artifacts may still be expected to provide additional “control information” and explanations in a prescribed format. This could unintentionally impose a higher operational burden on the most transparent projects, and may discourage publication or distribution to users in Japan.
What you might comment on
If you publish open-weight models, operate generative AI services reachable from Japan, or participate in dataset/crawling pipelines, you may want to comment on:
- Clearer scope boundaries and exclusions, so foundational tooling and community projects are not unintentionally captured.
- Proportionality for small developers, researchers, and community-led efforts.
- Stronger safeguards for trade secrets and practical limits on “control information” disclosure.
- A clearer and workable Open Source safe harbor that reduces uncertainty, not just a label.
Below are the official documents and the submission form.
Public comment guidance (JP) / includes deadline and says English submissions are accepted
https://public-comment.e-gov.go.jp/pcm/download?seqNo=0000304677
Principle-Code (draft) English PDF (provisional translation)
https://public-comment.e-gov.go.jp/pcm/download?seqNo=0000304679
Specific examples of disclosures (English PDF)
https://public-comment.e-gov.go.jp/pcm/download?seqNo=0000304681
Submission form (JP UI, English text accepted)
Thank you for taking a look, and for sharing this with relevant Open Source projects and providers.
Chairman of Open Source Group Japan
Shuji Sado