AI Guidelines for Business (Japan)

What it is

The AI Guidelines for Business (AI事業者ガイドライン) are Japan’s unified guidance on AI governance for organisations. The Ministry of Economy, Trade and Industry (METI) and the Ministry of Internal Affairs and Communications (MIC) first published them in April 2024. They replaced three earlier documents, MIC’s draft AI R&D Guidelines and AI Utilization Guidelines, and METI’s Governance Guidelines for Implementation of AI Principles.

The guidelines are deliberately soft law. They take a goal-based approach because, in the government’s words, rule-based regulation that sets detailed obligations might inhibit innovation, and law moves more slowly than AI technology. They were drafted with academic and research institutions, civil society and industry rather than by government alone.

They are a living document. Ver 1.1 followed in March 2025 and Ver 1.2 on 31 March 2026. The English text is a provisional translation, and the Japanese original prevails.

Who it applies to

The guidelines address business operators that develop, provide or use AI, including public bodies such as government agencies and municipalities. They group these into three AI business actors.

• AI developers build AI models and systems, including data collection, preprocessing and training, and keep models performing through post-training after deployment.
• AI providers incorporate AI systems into applications, products, existing systems or business processes, and provide them to AI business users and sometimes to consumers.
• AI business users use AI systems or services in their business, in the way the provider intended, and share information such as changes in the environment back with the provider.

One organisation can take on more than one of these roles. A company that fine-tunes a model and also uses it internally is both a developer and a business user. Individuals using AI outside a business are not the target.

What it requires

The main text is organised as basic philosophies (why), guiding principles (what) and, in the appendix, implementation approaches (how).

Ten common guiding principles

Every AI business actor is expected to work on ten common guiding principles.

1. Human-centric
2. Safety
3. Fairness
4. Privacy protection
5. Ensuring security
6. Transparency
7. Accountability
8. Education and literacy
9. Ensuring fair competition
10. Innovation

The first seven are about each organisation’s own conduct. Education and literacy, fair competition and innovation ask organisations to contribute to the wider society. Under accountability, the guidelines expect an organisation to appoint someone responsible for executing its accountability.

Actor-specific matters

Parts 3 to 5 set out additional matters for developers, providers and business users, organised by guiding principle. A business user, for example, is expected to use AI within the scope the provider set, avoid improperly inputting personal data or confidential information, and watch for bias in its inputs and prompts. Where AI output is used to evaluate a specific person or group, the business user should tell them AI is being used, have a human make the final judgement with automation bias in mind, and give an explanation when asked.

Building AI governance

Part 2 E asks organisations to build AI governance that keeps risks at levels acceptable to stakeholders. It recommends agile governance, a continuous cycle of environment and risk analysis, goal setting, system design, operation and evaluation, rather than fixed rules and procedures. The guidelines take a risk-based approach, so the measures should match the severity and probability of harm and the organisation’s resources.

Hiroshima AI Process

The guidelines incorporate the Hiroshima Process International Guiding Principles for All AI Actors, and the appendix reproduces the International Code of Conduct for organisations developing advanced AI systems.

AI agents and physical AI

Ver 1.2 defines an AI agent as an AI system that perceives its environment and acts autonomously to achieve a specific goal. It describes agentic AI as a goal-driven system that autonomously makes decisions and takes actions using multiple AI agents. It also defines physical AI, meaning systems that sense the physical world and turn decisions into physical actions.

The appendix

The appendix explains how to put the principles into practice, with behavioural goals and practical examples for building governance, and detailed guidance for each actor. It points to separate checklists and worksheets (published in Japanese only) that organisations can adapt to their own business and policies.

Key dates

• 19 April 2024. Ver 1.0 published.
• 28 March 2025. Ver 1.1 published.
• 1 September 2025. The AI Promotion Act came fully into force. Ver 1.2 refers to it.
• 31 March 2026. Ver 1.2 published by MIC and METI.

How it relates to other frameworks

The guidelines sit beneath the AI Promotion Act, which gives the government a statutory basis for guidance on appropriate AI development and use. They do not replace binding law. Personal information handling is still governed by the Act on the Protection of Personal Information.

Internationally, the guidelines draw on the OECD and G7 work that Japan led, including the Hiroshima AI Process. They reference the NIST AI Risk Management Framework and the EU AI Act when discussing concepts such as transparency. The governance cycle they describe fits naturally with an ISO/IEC 42001 management system, and the three actor roles broadly map to the provider and deployer split in the EU AI Act.

For agent programmes, Ver 1.2’s definitions give Japanese organisations a shared vocabulary for recording what an agent does, which systems it acts on and who oversees it.

A note on this page

This is Aicura’s reading of the guidelines, written to help you find your way around them. It is not legal advice and it is not the guidelines themselves. Read the primary text from MIC and METI, and take advice on how it applies to your organisation.