What do Japan's AI Guidelines for Business expect for AI literacy and training?

The AI Guidelines for Business (AI Jigyōsha Gaidorain) are Japan's unified, non-binding guidance on how businesses should develop, provide and use AI safely. They are published jointly by the Ministry of Internal Affairs and Communications (MIC) and the Ministry of Economy, Trade and Industry (METI), and they replaced three earlier sets of AI guidance when version 1.0 was issued in April 2024.

The Guidelines sort businesses into three roles: AI developers (who build AI models and systems), AI providers (who build AI into products, services or business processes and offer them to others) and AI business users (who use AI systems or services in their business). One company can hold more than one role. People who use AI outside business activities, and people affected by AI without using it, are outside the Guidelines' scope, but businesses are expected to consider them.

The Guidelines are a living document. The MIC publication page lists four versions to date, and version 1.2 is the one in effect as at September 2026. The main text sets out the basic philosophies and guiding principles (the what and why), and a much longer Appendix gives implementation methods and practical examples (the how).

Common guiding principle 8, Education/literacy (kyōiku, riterashī), asks each business to give its own people enough AI education to use AI correctly and to educate stakeholders about AI's risks. It sits with fair competition and innovation in a group of principles the Guidelines describe as activities expected in cooperation with society.

The core wording in version 1.2 of the Guidelines is: "Each AI business actor is expected to provide the persons engaged in AI in the AI business actor with the necessary education to gain the knowledge, literacy, and ethical views to correctly understand and use AI in a socially correct manner." The same paragraph expects education for stakeholders that considers AI's complexity, the misinformation it may produce and the possibility of intentional misuse.

The principle has three parts:

1. Ensuring AI literacy: take the necessary steps so that people engaged in AI reach "AI literacy of the level sufficient for the engagement". The standard is proportionate to each person's involvement, not a single required level.
2. Education and reskilling: because generative AI changes the division of tasks between people and AI, discuss education and reskilling actively to support new ways of working, and provide learning opportunities that account for differences in knowledge and skills between generations.
3. Support for stakeholders: provide stakeholders with education and literacy support as necessary to improve the safety of the whole AI system or service.

Table 1 of the Guidelines, which lists extra matters for each role, shows a dash against Education/literacy for developers, providers and business users. The Guidelines explain that a dash means each actor is expected to carry out the common principle itself, "rather than doing nothing". In other words, the literacy expectation applies to every role equally.

Two other principles shape training content. Principle 3, Fairness, suggests human intervention in AI decisions where needed and says it is essential to prevent human judgment from being influenced by automation bias. Principle 7, Accountability, notes that documentation can be kept with any suitable tool, as long as records are easy to check later. Both point toward training that builds judgment, not just awareness, as described in the human in the loop definition.

The Appendix to version 1.2 turns principle 8 into specific methods for each role in Appendices 3, 4 and 5. They are examples, not a mandatory syllabus, but they are the most detailed statement from the Japanese government of what workplace AI literacy should cover.

• AI developers (Appendix 3): education for developers that builds a mindset willing to change, including awareness of the latest attack methods, cooperation across the value chain and its trade-offs, and social responsibility. Developers are also expected to educate providers, business users and non-business users on proper use and potential risks.
• AI providers (Appendix 4): formulate AI policies that clarify roles and responsibilities and keep people involved in AI informed of them, of the characteriztics of reliable AI, of applicable laws and regulations and of potential adverse effects. Provide training that covers both the technological and socio-technological aspects of AI risk management.
• AI business users (Appendix 5): literacy content should cover basic knowledge of AI, mathematics and data science; bias in data and in use; issues of accuracy, fairness, privacy, security and the limits of AI; combining AI with skills such as asking questions and testing hypotheses; recognizing that AI generated false information circulates in media; and prompt writing for generative AI.

Appendix 5 also ties literacy to human oversight. Where people are meant to make the final decision on AI output, business users should acquire the ability and knowledge to assess that output appropriately. For training teams, this is the clearest link between literacy and a measurable competence.

Behavioral Goal 3-2 in Appendix 2 asks businesses to improve AI literacy strategically, under management's leadership, so that the AI management system runs properly. It is one of the behavioral goals for building AI governance, under the system design stage.

The Appendix gives a layered example: general literacy on AI ethics and AI reliability for officers, the management team and those responsible for the legal and ethical aspects of AI; training on AI technology, including generative AI, alongside AI ethics for staff on AI projects; and education on the positioning and importance of the AI management system for all employees. Its practice guidelines add:

• use training materials suited to job titles, duties and each member's role, including external lecturers and materials;
• make efforts to require all employees to receive education on AI ethics;
• provide training on the reliability of generative AI technology and its output, based on recent trends;
• define the personnel and skills needed to run the AI management system and specify educational content;
• use case studies and good practice from AI associations in internal training.

The Appendix also warns about a mismatch between employee development and the speed of technological change, because the AI literacy people need changes as the technology does. Behavioral Goal 1-3 asks businesses to evaluate their own AI proficiency, including employees' literacy in AI technology and ethics, and to reevaluate it at an appropriate time. That makes a periodic literacy baseline a natural first step for a training program.

No. The Act on Promotion of Research and Development, and Utilization of AI-related Technology (the AI Promotion Act, called the AI hō in Japanese government materials) is a framework law. It sets duties for the national government and general responsibilities for others, and it contains no provision requiring businesses to train staff and no penalties.

The Act is Act No. 53 of 2025. According to the Cabinet Office, it was promulgated and partly brought into force on June 4, 2025, and came fully into force on September 1, 2025, when the provisions on the AI Basic Plan and the Artificial Intelligence Strategic Headquarters took effect. The Japanese text on e-Gov is the authoritative version; an English translation is available for reference.

The Japanese text of Article 7 is framed differently for its two parts: businesses are to "endeavor" (tsutomeru) to use AI actively, and "must" cooperate with government measures. The cooperation duty has no defined content and no sanction. The Act also has no penalty provisions at all, which is a deliberate design choice rather than a gap.

Both documents reinforce literacy as a government priority and as part of good corporate AI governance, but neither creates a binding training requirement for businesses.

The Guideline for Ensuring the Appropriateness of Research and Development and Utilization of AI-related Technology, decided by the Artificial Intelligence Strategic Headquarters on December 19, 2025 under Article 13, aims to "encourage voluntary and proactive efforts" by all stakeholders. It asks utilization business operators, which it says include overseas business operators, to establish, operate and continuously improve AI governance, and it lists the implementation of education and training among the elements of that governance. It points to international norms and standards such as ISO/IEC 42001 as reference points.

The AI Basic Plan required by Article 18 was first adopted by Cabinet decision on December 23, 2025. A second plan was adopted on July 14, 2026 and is the current one, according to the Cabinet Office AI Basic Plan page. The provisional English translation of the second plan calls for "organizational management responses" in which organizations "improve literacy across the organization as a whole", promotes reskilling, and sets out human resource profiles. The first profile, professionals capable of appropriate AI utilization, is defined as people with "a minimum level of literacy in basic AI-related knowledge and insights" and is described as a goal all citizens should strive for.

Japan chose goal based soft law because it judged that detailed, rule based obligations could inhibit innovation and would lag the pace of AI development. The Guidelines say so directly: they were drawn up on "the goal-based concept that would lead to the achievement of purposes through soft laws without any legally binding force".

The approach relies on agile governance: businesses analyze their own risks, set goals, design and run an AI management system, evaluate it and repeat the cycle as conditions change. Government sets principles and publishes guidance, and existing laws, such as privacy and intellectual property law, continue to apply. For AI literacy this means the content and depth of training are for each business to decide, scaled to the risk of its AI use.

Soft law still has consequences. The Guidelines warn that if society considers a business's efforts inappropriate or insufficient, it may lose business opportunities and find it hard to maintain business value. Binding laws also still apply to AI use: the Guidelines tell businesses to observe existing laws such as the Act on the Protection of Personal Information (see APPI employee training).

Start from the roles your organization holds under the Guidelines, then map each expectation to a learning outcome and to the evidence that shows it was achieved. The table below is an illustrative mapping built from principle 8 and the Appendix; it is not an official template.

The method follows how to map training to compliance obligations. If the program uses generative AI to draft its own materials, apply the review controls in how to govern AI generated learning content and how to use AI to create training content safely.

Because nothing is mandatory, evidence serves governance and disclosure rather than a regulator's inspection. A practical record set is:

1. An AI literacy baseline for the organization, repeated periodically (Behavioral Goal 1-3).
2. A role map showing which people are engaged in AI as developers, providers or business users.
3. Versioned training content linked to the AI policy it teaches.
4. Completion, assessment and attestation records kept as an audit trail.
5. A review log showing when content was updated after a Guidelines revision or a change in AI tools.

Knowledge Foundry models principle 8 and the relevant Appendix methods as expectations, links each to the organization's own AI policy and roles, and derives learning outcomes and assessment points before any content is written. When MIC and METI publish the next version of the Guidelines, the changed expectations can be traced to the outcomes and materials they affect. The same framework can hold EU AI Act Article 4 alongside the Japanese Guidelines for multinational programs.