Japan AI Regulation & Governance

Japan's Act on Promotion of Research, Development and Utilization of Artificial Intelligence-Related Technologies (the AI Promotion Act, Act No. 53 of 2025; e-Gov law ID 507AC0000000053; also recorded as JP-Act-53-2025) is the jurisdiction's framework AI statute. Most of its provisions took effect on 4 June 2025, with Chapters 3 and 4 - which establish the AI Strategy Headquarters and its organs - and certain Supplementary Provisions articles deferred to 1 September 2025; available reporting places the Act's full commencement at 1 September 2025. A retrieved official translation confirms that Article 9 of the AI Promotion Act imposes a coordination duty on the national government - requiring it to take measures strengthening coordination among the national government, a local government, a research and development institute and a utilization business operator - rather than a direct obligation on firms. Article 7 of the AI Promotion Act separately imposes an efforts duty on AI developers, providers and deployers to cooperate with government AI-related measures, carrying no penalty for non-compliance. Taken together, the Act is a promotional framework statute that establishes the AI Strategy Headquarters and operates through non-binding guidelines rather than directly enforceable obligations, with no licensing or registration regime for AI (CLM-JP-7dc45984; CLM-JP-02bf841f). Acting under Article 13 of the AI Promotion Act, the AI Strategy Headquarters published non-binding AI Utilization Guidelines on 19 December 2025, recommending a risk-based approach for all subjects of the Act. Under Article 18(1) of the AI Promotion Act, Japan adopted its first AI Basic Plan on 23 December 2025 and a second AI Basic Plan on 14 July 2026. The AI Strategy Headquarters is the body tasked with AI policy coordination, AI Basic Plan formulation and issuance of the AI Utilization Guidelines. An amending enactment to the AI Promotion Act creates an election AI-content labelling duty that has not yet commenced: it is scheduled to take effect on 1 March 2027; the substantive content of that duty is addressed in the Transparency & Synthetic Content module.

Japan does not operate a dedicated binding high-risk AI system classification regime. The applicable standing obligation on operators is the Article 7 efforts duty under the AI Promotion Act (Act No. 53 of 2025), which requires AI developers, providers and deployers to cooperate with government AI-related measures without attaching a penalty for non-compliance. Supervisory expectations for high-autonomy AI operation are set out in the METI/MIC AI Guidelines for Business (AI事業者ガイドライン), now at version 1.2, effective 31 March 2026. Because AI agents and physical AI act autonomously, the guideline states that human monitoring alone may be insufficient to keep pace with high-speed AI-to-AI interaction, and it recommends that AI providers build mechanisms allowing human judgement to intervene in the operation of AI agents and physical AI (CLM-JP-0fc5f9f9; CLM-JP-2406b456). This guidance is non-binding: it sets a recommended operator practice rather than a directly enforceable rule, and version 1.2's addition of an annex on AI agents is the most recent development in the standing guidance.

Since 23 July 2023, Japan has required a prior individual export licence from the Ministry of Economy, Trade and Industry (METI) for 23 items of advanced-semiconductor manufacturing equipment, except where destined for 42 listed friendly destinations; this control bears on the supply chain underlying general-purpose and foundation-model-capable hardware. On the guidance side, the Japan AI Safety Institute (J-AISI) issued a voluntary Guide to Red Teaming Methodology on AI Safety, now at version 1.10 (March 2025), setting non-binding considerations for adversarial testing of AI systems - including general-purpose systems - both before and after release. Neither instrument creates a dedicated binding compliance regime specific to general-purpose or foundation models beyond these export controls and the voluntary testing guidance.

No binding synthetic-content labelling duty is currently in force in Japan. Amendments to the Public Offices Election Act and the Information Distribution Platform Act were enacted on 13 July 2026 and will take effect on 1 March 2027; from that date, AI-generated or AI-altered election campaign images and videos that could be mistaken for unaltered footage must be clearly labelled as AI-created or altered (CLM-JP-e8862054; CLM-JP-97953e93). In the interim, the AI Strategy Headquarters' AI Utilization Guidelines, issued under Article 13 of the AI Promotion Act, direct businesses to strive to develop technologies capable of determining whether content is AI-generated - such as digital watermarks, provenance management and related APIs - as necessary; this is a non-binding direction rather than a mandatory technical standard.

Article 30-4 of Japan's Copyright Act permits use of copyrighted works for AI-training purposes without the right holder's consent, provided the purpose is not to enable enjoyment of the thoughts or sentiments expressed in the work, subject to statutory exceptions. On the question of copyrightability of AI-generated output, the Agency for Cultural Affairs' (Bunkacho) non-binding guidance states that an AI-generated output is copyrightable only where a human shows both creative intent and creative contribution to the output. Refining that standard, Japan's Copyright Subcommittee has articulated a test that weighs the volume of prompting, the number of generation attempts and the degree of human selection involved, with no single factor alone sufficient to establish creative contribution.

Two binding instruments currently govern sector-specific or public-sector AI use in Japan. First, since 23 July 2023, a Foreign Exchange and Foreign Trade Act (FEFTA) Ministerial Order amendment has required a prior individual METI export licence for 23 items of advanced-semiconductor manufacturing equipment, except to 42 listed friendly destinations. Second, Digital Society Promotion Standard Guideline DS-920, in force since 27 May 2025, binds national ministries and agencies: under section 6.2, section 3.2 and Annex 3 of Digital Society Promotion Standard Guideline DS-920, each ministry or agency must appoint a Chief AI Officer (CAIO) who applies a four-axis test to determine whether a given generative-AI use is high-risk, with procurement required to apply a standardised checklist. Digital Society Promotion Standard Guideline DS-920 version 2.0 is scheduled to take effect on 1 September 2026 and will have the CAIO govern generative-AI use on the same four-axis basis while applying an expanded standardised procurement checklist of 21 evaluation items. On the supervisory-expectations tier, the Financial Services Agency (FSA) issued an AI Discussion Paper (version 1.0, March 2025, updated to version 1.1 in March 2026) setting non-binding supervisory expectations for financial institutions' use of AI in credit screening and underwriting, with a focus on explainability and fairness (CLM-JP-dfda187c; CLM-JP-b4bf2842).

Japan has no dedicated binding incident-reporting statute for AI; the operative mechanism sits within the AI Promotion Act's investigatory and disclosure powers, administered by the AI Strategy Headquarters. Where the AI Strategy Headquarters investigates a serious AI-related incident and a business does not follow the guidance that results, the government may publicly disclose that entity's details. The AI Strategy Headquarters is the body responsible for AI policy coordination, AI Basic Plan formulation and the AI Utilization Guidelines, and exercises this disclosure power within that remit.

The AI Promotion Act fixes no fine, criminal offence or administrative monetary penalty for breach of its provisions; its only enforcement lever is investigation, guidance and public disclosure. In practice this operates through the AI Strategy Headquarters: where that body investigates a serious AI-related incident and a business does not follow the resulting guidance, the government may publicly disclose that entity's details. The AI Strategy Headquarters is the central authority for AI policy coordination, AI Basic Plan formulation and the AI Utilization Guidelines, and currently exercises reputational rather than financial enforcement powers.

Japan's regulatory sandbox regime was established in 2018 under the Act on Special Measures for Productivity Improvement (in force from 6 June 2018) and was made permanent in 2021, with its sandbox provisions carried into the Act on Strengthening Industrial Competitiveness. Under this regime, a business holding a ministerially-certified implementation plan may run demonstration tests of new technologies or business models - including AI and IoT - outside otherwise-applicable regulation, for a certified participant group and for a limited period (CLM-JP-37e71a4d; CLM-JP-dc3ca382).