Japan AI Promotion Act (Act on the Promotion of Research, Development and Utilization of AI-Related Technologies) — provisions explorer

International Coordination

Act No. 53 of 2025, Arts. 17 & 3(5)

Editor rationale: Art. 17 is an explicit operative directive that the State promote international cooperation and actively participate in international AI norm-setting, reinforced by the Basic Philosophy principle of leadership in international cooperation (Art. 3(5)). Directly on point — governs, medium confidence (still aspirational, no concrete obligations).

Development-Rights Framings

Act No. 53 of 2025, Arts. 1 & 3(3)

Editor rationale: The whole Act is an innovation-first PROMOTION statute: Art. 1's purpose is to promote AI R&D and use for economic/social development, and Art. 3(3) makes comprehensive promotion across all stages a governing principle. The development/promotion-as-priority framing is the explicit operative spine of the law — governs, medium confidence.

Transparency Obligations

Act No. 53 of 2025, Art. 3(4)

Editor rationale: Art. 3(4) makes 'securing transparency in the processes' of AI R&D and use an explicit operative principle, tied to preventing harm to citizens' rights. It is a statutory principle (governs), but a non-binding basic-law aspiration with no penalty, definitional threshold, or disclosure mandate — hence low confidence.

Compute-Threshold Reporting

Act No. 53 of 2025, Art. 12

Editor rationale: Art. 12 addresses large-scale compute, but as State-provided SHARED INFRASTRUCTURE to be built, not as reporting or threshold-based disclosure of compute by developers. It touches the compute domain only indirectly and in the opposite (enabling) direction — implicit, low confidence. Not a compute-reporting obligation.

Training-Data Rights

Act No. 53 of 2025, Arts. 12 & 3(4)

Editor rationale: Art. 12 directs the State to develop and promote shared use of datasets/intellectual infrastructure for AI (enabling, not regulating training data); Art. 3(4) flags personal-data leakage and copyright infringement as harms to guard against. Neither imposes a training-data rule on developers, so this is indirect support at most — implicit, low confidence.

Individual Redress

Act No. 53 of 2025, Art. 16

Editor rationale: The Act creates no individual remedy, complaint right, or cause of action. Art. 16 only has the State analyze rights-infringement cases and issue guidance/advice (a state-level soft response). At most an indirect, systemic gesture toward addressing harms, not a redress mechanism — downgraded to implicit, low confidence.

Foundation Models / GPAI

Act No. 53 of 2025, Arts. 2 & 12

Editor rationale: The Act's broad definition (Art. 2) covers foundation/large models, and Art. 12's large-scale-compute and dataset infrastructure is most relevant to large-model training. But the Act nowhere singles out foundation/frontier models, applies no model-tier obligations, and is purely promotional — so any coverage is indirect. Implicit, low confidence.

Sovereign AI Doctrine

Act No. 53 of 2025, Art. 3(2)

Editor rationale: Art. 3(2) frames AI as foundational to socioeconomic development AND national security, with the aim of preserving Japan's domestic R&D capacity and industrial competitiveness — a sovereign-capability framing. It is a principle/aim, not an operative build-out mandate, so implicit, low confidence.

Technological Sovereignty

Act No. 53 of 2025, Art. 3(2)

Editor rationale: Same Art. 3(2) language casts AI as nationally strategic — maintaining domestic R&D capacity and competitiveness, framed partly through a national-security lens. A sovereignty/strategic-autonomy framing principle, not an operative control (no FDI screening or export rule), so implicit and low confidence.

National Security Carveouts in AI Regulation

Act No. 53 of 2025, Art. 3(2)

Editor rationale: Art. 3(2) explicitly names national security as a reason AI is strategically important, but the Act creates no security-specific exemption, carve-out, or special regime (and has no penalties to carve out from). National security appears only as a promotional rationale, so this is indirect — implicit, low confidence.