Artificial Intelligence Law in Japan

The National AI Strategy: Society 5.0

Japan’s approach to artificial intelligence law is inseparable from the Society 5.0 (Sōshiete 5.0) framework, first articulated in the Fifth Science and Technology Basic Plan (2016–2020). Society 5.0 envisions a “super-smart society” in which cyber-physical systems integrate to solve social challenges — an aging population, labor shortages, regional depopulation, and natural disaster resilience — through AI, the Internet of Things (IoT), and big data. Unlike the Industry 4.0 paradigm (which focuses on manufacturing), Society 5.0 posits AI as a comprehensive social infrastructure.

The AI Strategy Council (AI Senryaku Kaigi), established in 2019 under the Cabinet Office, coordinates AI policy across ministries. Its 2019 report, “AI Strategy 2019,” and the subsequent “AI Strategy 2022” set out three pillars: the promotion of AI R&D; the development of AI-ready human resources; and the establishment of a governance framework for responsible AI. This strategy is operationalized through the Integrated Innovation Strategy Promotion Council and the Council for the Promotion of Regulatory Reform, which periodically issue action plans for AI deployment across healthcare, mobility, finance, and manufacturing.

The Basic Act on the Advancement of AI

Japan enacted the Basic Act on the Advancement of AI (AI Suishin Kihon Hō, Law No. XX of 2024) to provide a statutory foundation for national AI policy. The Act establishes the basic principles for AI development and use, including: (i) respect for fundamental human rights; (ii) the maintenance of democracy and the rule of law; (iii) the promotion of human-centric AI; (iv) transparency and explainability; and (v) international harmonization. It mandates the formulation of a Basic Plan on the Advancement of AI every five years and creates a Council on the Advancement of AI within the Cabinet Office to coordinate cross-ministerial policy.

The Basic Act does not impose binding obligations on private actors. Instead, it functions as a framework statute that authorizes sector-specific regulations and soft-law governance instruments. This approach reflects Japan’s preference for coordinated governance (renkei-gata gabanansu) over the prescriptive, risk-tiered model adopted by the European Union in the EU AI Act.

The Act on the Protection of Personal Information

The Act on the Protection of Personal Information (APPI) is the most immediately binding legal constraint on AI systems that process personal data. Enacted in 2003 and substantially amended in 2015, 2020, and 2022, the APPI regulates the collection, use, and transfer of personal information (kojin jōhō) by businesses. The 2020 amendments, effective April 2022, introduced provisions of particular relevance to AI: (i) mandatory reporting of data breaches to the Personal Information Protection Commission (PPC) ; (ii) restrictions on the provision of personal data to third parties in foreign jurisdictions; (iii) the prohibition of “inappropriate use” of personal data; and (iv) an expanded definition of “personal data” to include biometric data and behavioral profiles.

For AI developers, the key APPI requirements are the purpose limitation principle (Article 16), which requires that personal data be used only within the scope of purposes notified to the data subject, and the opt-in consent requirement for the provision of data to third parties (Article 23). The PPC’s guidelines on AI and big data emphasize that the use of personal data for machine learning model training must either fall within the originally notified purpose or be based on separate consent, unless the data has been sufficiently anonymized (processed into “anonymized processed information” under Article 36).

Sectoral AI Regulation

Japan has adopted sector-specific AI regulation in domains where safety and liability concerns are acute. Under the Road Traffic Act (Dōro Kōtsū Hō), amended in 2019 and 2022, Level 3 and Level 4 automated driving systems are subject to a licensing regime. The 2019 amendments permit “automated driving equipment” (jidō unten sōchi) that meets technical safety standards, with the vehicle owner bearing liability for traffic violations, although criminal liability for accidents remains with the human user. The 2022 amendments introduced a permit system for Level 4 automated mobility services in designated areas, administered by the Ministry of Land, Infrastructure, Transport and Tourism (MLIT) .

In healthcare, the Pharmaceuticals and Medical Devices Act (PMD Act) and its associated regulations govern AI-powered medical devices (jinkō chinō o mochiita iryō kiki). The Pharmaceuticals and Medical Devices Agency (PMDA) has issued guidelines requiring that software as a medical device (SaMD) incorporating machine learning obtain pre-market approval or certification, with post-market surveillance obligations to monitor algorithmic drift. The PMDA’s “Guidelines on Artificial Intelligence-Based Medical Devices” (2021) require developers to demonstrate the reproducibility and robustness of AI models across diverse patient populations.

AI Tort Liability Under the Civil Code

In the absence of a specific AI liability statute, claims for AI-caused harm fall under the general tort provisions of the Civil Code (Minpō). Article 709 imposes liability on a person who “intentionally or negligently” infringes another’s rights or legally protected interests. For AI system failures, the plaintiff must establish that the operator, developer, or deployer breached a duty of care. Japanese courts have not yet decided a leading case on AI tort liability, but academic commentary suggests that liability will typically fall on the operator (shiyōsha) as the person best positioned to control the risk and most likely to be insured.

Strict liability provisions may apply in certain contexts. Article 717 of the Civil Code imposes strict liability on the possessor of a “structure on land” for defects in construction or maintenance. If an AI system is embedded in a physical structure (such as an elevator or autonomous vehicle), the structure owner may be strictly liable. Similarly, the Product Liability Act (Seizōbutsu Sekinin Hō, 1994) imposes strict liability on manufacturers for defects in products, which applies to AI-integrated products if the software defect renders the product unsafe.

AI Governance Guidelines

The Ministry of Economy, Trade and Industry (METI) has been the principal architect of soft-law AI governance in Japan. Its “AI R&D Guidelines” (2017, revised 2019) set out ethical principles for AI developers: (i) benefit to humanity; (ii) fairness and non-discrimination; (iii) transparency and explainability; (iv) controllability; and (v) safety. These are supplemented by the “AI Utilization Guidelines” (2019), which address standards for AI deployment by businesses, and the “Governance Guidelines for Implementation of AI Principles” (2021, revised 2022), which provide a framework for internal corporate AI governance modeled on the “comply-or-explain” approach.

In 2023, METI and the Cabinet Office jointly published the “Proposal for the Ideal Approach to AI Governance”, which moved toward a risk-based framework distinguishing between “specified AI” (high-risk) and other AI, drawing on the EU approach while rejecting mandatory third-party conformity assessment for all high-risk systems. The proposals informed the Basic Act on the Advancement of AI and contemplate voluntary certification systems administered by private-sector AI auditing bodies.

Comparative Perspectives

Japan’s AI governance model shares the EU’s commitment to human-centric AI but diverges in its preference for soft law, sectoral regulation, and non-binding principles over prescriptive regulation. While the EU AI Act imposes binding obligations on providers and deployers of high-risk AI systems, with fines up to 7% of global annual turnover, Japan relies on guidelines supplemented by sectoral statutes (APPI, Road Traffic Act, PMD Act). The US Executive Order on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence (October 2023) similarly emphasizes existing statutory authority and voluntary commitments, aligning more closely with Japan’s approach.

Japan has been active in international AI governance forums, including the G7 Hiroshima AI Process (2023), which produced the International Code of Conduct for Advanced AI Systems, and the OECD AI Policy Observatory. The Japan-EU Digital Partnership and the US-Japan Competitiveness and Resilience (CoRe) Partnership include AI governance as a pillar, reflecting Japan’s strategy of regulatory alignment without wholesale adoption of the EU model.

Conclusion

Japanese AI law is characterized by a coordinated governance approach that combines a framework statute (the Basic Act on the Advancement of AI), sector-specific regulation in safety-critical domains, and a dense network of non-binding guidelines. The APPI provides the primary binding constraint on data-driven AI, while tort law fills gaps in liability for AI-caused harm. Japan’s model reflects its institutional preference for incremental, consensus-based regulation and its strategic goal of promoting AI innovation within the Society 5.0 vision. Whether this approach proves adequate as AI capabilities advance — particularly for general-purpose AI and generative foundation models — remains an open question.