What it is
This guideline was decided by Japan’s AI Strategic Headquarters on 19 December 2025 under Article 13 of the AI Promotion Act, after a public comment period from 5 to 11 December 2025. It is the government-wide reference point for appropriate AI development and use, and the hub for ministry guidelines listed on the same Cabinet Office page.
It aims at voluntary and proactive efforts and has no penalties, though some wording is firm (organizations are “required to respond appropriately”).
What it contains
Section 1: shared foundations. Ten elements (human-centricity, fairness, safety, transparency, accountability, security, privacy, fair competition, AI literacy, innovation) and four policies: a risk-based approach, stakeholder involvement, lifecycle governance, and agile PDCA cycles.
Section 2: developers and providers.
- Lifecycle AI governance, referencing ISO/IEC 42001 and the Hiroshima Process Reporting Framework.
- Transparency.
- Safety: hallucination, bias, and deepfakes, with digital watermarking encouraged.
- Business continuity.
- Consideration for data and intellectual property holders.
- Red teaming, suggested in footnotes.
Section 3: government. Each ministry appoints an AI governance officer.
Section 4: citizens. Check sources and keep final decisions with humans.
Where to find it
The Cabinet Office AI guideline page carries the Japanese text, the public comment results, and provisional English versions (full text dated 16 Jan 2026, summary dated 20 Feb 2026). The Japanese text governs. The page also lists sector guidelines, such as MIC’s AI security guideline (March 2026).
Reuse terms
Cabinet Office content is reusable under PDL1.0 unless otherwise noted, with source attribution and an edit statement; logos and character designs are excluded. Check each attachment and linked ministry guideline separately, because each ministry has its own terms page.
Using the guideline in AI and RAG
- Index the Japanese text as authoritative and the English versions as provisional, each with its date.
- Link each sector guideline as a child document with its own issuer, date, and terms.
- Tag Section 2 items by topic (governance, transparency, safety, IP) for gap checks.
- Record the ISO/IEC 42001 reference without loading ISO text, which is licensed separately.
- Add PDL1.0 attribution to outputs.
Common mistakes
- Confusing this guideline with METI and MIC’s AI Guidelines for Business. They are separate documents with different issuers.
- Assuming the English version is authoritative.
- Treating linked ministry guidelines as having the same terms or dates.
- Reading “required” wording as creating fines.
Checklist
- Japanese text and provisional English stored with dates
- Sector guidelines linked separately
- Section 2 topics tagged
- ISO references noted, ISO text excluded
- Attribution in outputs
How AskDona helps
AskDona maps your AI governance framework to the guideline’s ten elements and Section 2 expectations, citing the specific item behind each finding.
Run a sample Article 13 guideline check
AskDona shows the evidence; your AI governance team decides.
Editor’s note. This article is general information, not legal advice. The “conditions to check” rating comes from our source list and is a first-pass judgment, not a legal opinion.
FAQ
When was the Article 13 guideline issued? 19 December 2025, by the AI Strategic Headquarters.
Is it binding? It seeks voluntary efforts and has no penalties.
Is there an English version? Yes, provisional English versions dated January and February 2026.
How does it relate to other guidelines? It serves as the hub; ministry guidelines such as MIC’s AI security guideline are listed with it.
References
All pages opened and checked by October 26, 2026.
| Source | What it supports |
|---|---|
| Cabinet Office: AI guideline page | Decision date, public comment, English PDFs, sector guidelines |
| Cabinet Office: Guideline, provisional English (PDF) | Article 13 basis, structure, voluntary nature |
| Cabinet Office: Terms of use | PDL1.0 |