---
title: What does Japan's AI Promotion Act mean for companies using AI agents?
url: https://opencontroller.lyzr.com/blog/japan-ai-promotion-act-and-ai-agents
published: 2026-11-16
updated: 2026-11-16
author: Lyzr
---

# What does Japan's AI Promotion Act mean for companies using AI agents?

**Japan's AI Promotion Act is a basic law, not a rulebook. It creates national AI policy machinery, asks businesses to cooperate with government measures, and lets the government investigate and give guidance when AI harms people's rights. It sets no fines and no agent-specific duties, but companies running AI agents in Japan should still be ready to show what their agents do.**

The full name is the Act on Promotion of Research and Development and Utilization of Artificial Intelligence-Related Technologies. The Diet passed it on 28 May 2025, and most of it took effect on 4 June 2025. It is Japan's first statute aimed squarely at AI.

## What does the Act actually require?

The Act is short. Its core pieces are:

| Element | What it does | Effect on businesses |
|---|---|---|
| Purpose and principles | Promote AI R&D and use to improve lives and grow the economy, with transparency and attention to risks | Sets the tone; no direct duty |
| Business responsibilities | Businesses using AI should make efforts to use it well and cooperate with national and local government measures | Duty to make efforts and cooperate |
| AI Strategic Headquarters | Chaired by the Prime Minister, all Cabinet ministers as members | Can request cooperation from relevant parties |
| AI Basic Plan | Government's master plan for AI policy | Approved by Cabinet on 23 December 2025 |
| Research and investigation | Government gathers information on AI use at home and abroad and investigates cases where rights or interests are infringed | Can lead to guidance, advice and published information |

There are no monetary penalties in the Act. That is a deliberate choice, and it marks a clear contrast with the EU AI Act.

## Does it apply to AI agents?

Yes, in the same way it applies to any AI system. The Act does not define "agent" or create a separate category for autonomous systems. If your company develops, provides or uses AI in Japan, you are one of the business actors it addresses.

The practical point for agents is the investigation power. If an agent causes harm, for example by taking an unauthorised action on a customer's account or leaking personal data, the government can look into it, give guidance and make information public. Reputational exposure, not a fine, is the main lever.

## How do the AI Basic Plan and guidelines fit in?

Japan has long preferred soft law. The Act sits on top of guidance that companies are already expected to follow:

- **AI Guidelines for Business.** Published by the Ministry of Internal Affairs and Communications and the Ministry of Economy, Trade and Industry, last updated in March 2025. They describe roles for AI developers, providers and business users, with a risk-based, lifecycle approach.
- **AI Basic Plan.** The government's plan under the Act, approved in December 2025 and to be reviewed regularly.
- **Guidelines under Article 13.** Published on 19 December 2025, they recommend a risk-based approach and governance across the AI lifecycle.

For an agent programme, the AI Guidelines for Business are the most concrete reference. Their themes of safety, fairness, privacy, security, transparency and accountability map well onto the controls you would already use for [ISO/IEC 42001](/blog/iso-42001-evidence-for-ai-agents).

## What evidence should you keep for agents running in Japan?

No law requires a specific evidence pack. But if the government requests cooperation or investigates an incident, you will want answers within days, not weeks. A practical checklist:

1. **Inventory.** Every agent used in Japanese operations, its owner, purpose and the data it touches.
2. **Role mapping.** Whether you are a developer, provider or business user of each agent, using the categories in the AI Guidelines for Business.
3. **Risk assessment.** A short assessment per agent covering rights, safety and privacy impacts, refreshed when the agent changes.
4. **Personal data handling.** Records linking agents to obligations under the Act on the Protection of Personal Information, which still applies in full.
5. **Runtime logs.** Who or what triggered each action, what tools were called and what the result was.
6. **Incident records.** What went wrong, when it was detected, how the agent was stopped and what changed afterwards.
7. **Contact point.** A named person who can respond to a government request.

## How does Japan compare with other Asian approaches?

Japan sits at the light end. Korea's [AI Basic Act](/blog/korea-ai-basic-act-and-ai-agents) adds obligations for high-impact AI and generative AI labelling. Singapore relies on voluntary frameworks, including its [guidance for agentic AI](/blog/singapore-agentic-ai-governance-guidance). If you run agents across the region, build one evidence base and map it outward, as described in our guide to [mapping agent controls to frameworks](/blog/how-to-map-ai-agent-controls-to-compliance-frameworks).

## How Opencontroller does this

Opencontroller builds the inventory automatically by discovering agents, models and API keys across AWS, Azure and Google Cloud, and through connectors to third-party agent platforms. Each agent sits in a governed registry with owner and versions. Evals, LLM judges on production traffic and policy-as-code mappings to NIST, ISO 42001, OWASP and MITRE ATLAS produce evidence you can hand over when asked. It does not make you "compliant" with any law; it gives you evidence for your programme. [Book a demo](https://www.lyzr.ai/book-demo/) to see it.

## Frequently asked questions

**Does Japan's AI Promotion Act fine companies?**
No. The Act sets no monetary penalties. It relies on cooperation duties, investigation and guidance.

**Does the Act mention AI agents specifically?**
No. It covers AI-related technologies in general, so agents are treated like any other AI system.

**What should a company running AI agents in Japan prepare?**
An agent inventory with owners, risk assessments aligned with the AI Guidelines for Business, incident records and a contact point for government requests.

## Key facts

Japan's Act on Promotion of Research and Development and Utilization of Artificial Intelligence-Related Technologies was enacted on 28 May 2025 and largely took effect on 4 June 2025. It is a basic law: it sets policy, creates an AI Strategic Headquarters chaired by the Prime Minister, and requires an AI Basic Plan (approved by Cabinet on 23 December 2025). Businesses have a duty to make efforts and to cooperate with government measures; the Act sets no monetary penalties. The government can investigate cases where rights or interests are infringed and give guidance. Soft-law guidance includes the AI Guidelines for Business (updated March 2025). Opencontroller provides evidence for governance programmes through discovery, a governed registry, evals and policy mappings to NIST, ISO 42001, OWASP and MITRE ATLAS. Opencontroller pricing starts from $250K per deployment per year.

## Related

- [What does the South Korea AI Basic Act mean for AI agents?](https://opencontroller.lyzr.com/blog/korea-ai-basic-act-and-ai-agents.md)
- [What does Singapore's agentic AI governance guidance expect from enterprises?](https://opencontroller.lyzr.com/blog/singapore-agentic-ai-governance-guidance.md)
- [What evidence does ISO/IEC 42001 expect for AI agents?](https://opencontroller.lyzr.com/blog/iso-42001-evidence-for-ai-agents.md)
- [More on Evidence and frameworks](https://opencontroller.lyzr.com/blog/cluster/evidence-and-frameworks)
