At a glance
- Japan’s AI law is the Act on Promotion of Research and Development, and Utilization of Artificial Intelligence-related Technology (Act No. 53 of 2025), fully in force since September 1, 2025. It contains no penalty provisions and no licensing, risk-classification or labeling duty for companies.
- Under Article 7, a business that intends to develop or provide AI products or services, or to use AI in its business activities, must cooperate with the measures national and local governments implement under Articles 4 and 5, in accordance with the basic principles.
- Most AI-specific expectations for companies sit in guidelines that do not bind: the Article 13 appropriateness guideline (December 19, 2025) and the AI Guidelines for Business, version 1.2 (March 31, 2026). Binding rules reach AI through existing law, chiefly the Act on the Protection of Personal Information (APPI) and the Copyright Act.
- APPI amendment Act No. 56 of 2026, promulgated July 17, 2026, will make consent unnecessary, subject to conditions, for providing personal data to third parties, and for acquiring special care-required personal information that has been made public, where the data is used only to prepare statistical information, including AI development that can be classified as statistical compilation. It also adds surcharges for certain serious violations. In principle it takes effect on a date set by Cabinet Order within two years of promulgation.
Does Japan have an AI law?
Yes, but it promotes AI rather than regulating it. The AI Promotion Act (Act No. 53 of 2025) sets basic principles, assigns responsibilities, requires a government AI Basic Plan and creates the AI Strategic Headquarters under the Cabinet. It imposes no fines, licensing or labeling duties on companies. It requires businesses to cooperate with government measures and directs the government to analyze cases in which AI used for wrongful purposes or by inappropriate methods infringes people’s rights and interests, and to give guidance, advice and information.
The English title comes from the Ministry of Justice’s Japanese Law Translation database, whose translations “are not official texts”; only the Japanese text (人工知能関連技術の研究開発及び活用の推進に関する法律) has legal effect. The Cabinet Office calls it the AI Act (AI法); this guide says AI Promotion Act to keep it apart from the EU AI Act.
The Cabinet Office gives the enactment date as May 28, 2025. The Act was promulgated and partly took effect on June 4, 2025. Chapters III and IV, on the AI Basic Plan and the Headquarters, took effect by Cabinet Order on September 1, 2025. That day the Headquarters was set up, chaired by the Prime Minister. As of November 12, 2026, the e-Gov law database lists no amendment to the Act after that date.
What does the AI Promotion Act require of businesses?
Article 7 covers any person who intends to develop or provide AI products or services, or to use AI in its business activities (a “utilization business operator”). It “is to endeavor to enhance the efficiency and advancement of business activities and create new industries through active utilization of artificial intelligence-related technology,” and it “must cooperate with the measures implemented by the national government under the provisions of Article 4 and by a local government under the provisions of Article 5, in accordance with the basic principles.”
The Cabinet Office’s English overview, a provisional translation, says “shall endeavor to cooperate”, but its Japanese overview, like the Act’s Japanese text, says 協力しなければならない (our translation: “must cooperate”). The Act does not define cooperation and sets no sanction for failing to give it. It has no territorial-scope article, but the Article 13 guideline says the Article 7 term includes “overseas business operators.”
The government’s tools are these:
- Guidelines (Article 13). The government “is to establish guidelines in accordance with the purport of the international norms and take other necessary measures.”
- Research, guidance and advice (Article 16). The government collects information, analyzes cases where citizens’ rights and interests are infringed through AI used for wrongful purposes or by inappropriate methods, and, based on the results, “is to give guidance, advice to a research and development institute, a utilization business operator and other people, and provide information for them, or take other necessary measures.”
- Requests by the Headquarters (Article 25(2)). If the Headquarters finds it particularly necessary, it may ask for necessary cooperation from persons other than the relevant administrative bodies, local governments, incorporated administrative agencies, local incorporated administrative agencies and special corporations listed in Article 25(1).
When the Cabinet approved the bill on February 28, 2025, the minister in charge was asked why it carried no penalties. He said the government had chosen to avoid excessive regulation that would impede innovation (「イノベーションの阻害をする過剰な規制は避けることとした」, our translation). He also said the bill conferred no new powers (「本法案によって新たな権限が付与されるものではございません」). Where existing law cannot address a case, he said, the government would investigate and respond, for example by publishing the results (「調査結果を公表するなど」). Publication is a stated intention. Article 16 itself lists guidance, advice, information and “other necessary measures”.
On January 16, 2026, after Grok’s image editing had been used to alter and spread others’ photos without permission, including into sexual images, the Minister for Artificial Intelligence Strategy said the Cabinet Office had asked X Corp to improve promptly, had sent a written inquiry and was seeking a prompt report. If nothing improved, the minister said, the government would consider every option, including guidance under the AI Act and legislative measures. That was a request, not an order or a penalty.
What is in the AI Basic Plan?
The AI Basic Plan directs government work and places no duty on companies, but it shows where rules are heading. Under Article 18, the Headquarters drafts it and the Cabinet decides it. The Cabinet decided the first plan, “Japan Rebooted” through “Trustworthy AI”, on December 23, 2025, and the second, “Japan AI Transformation: A Stronger Nation. Greater Prosperity.”, on July 14, 2026. The English texts are provisional translations, and the plan is to be updated “annually for the foreseeable future.” The second plan commits the government to:
- Run Article 16 research “in a more agile manner” and “proactively and continuously review systems and related measures, including the AI Act.”
- “Take measures to enable business operators to appropriately establish their own complaint-handling systems.”
- Develop guidelines on how tort law treats infringement of publicity and other personality rights by unauthorized use of voices, and examine the need for legislative reforms, including amendments to the Unfair Competition Prevention Act.
- Develop subordinate regulations for the smooth enforcement of the 2026 APPI amendment and, centered on the AI Safety Institute, strengthen AI model evaluation capabilities.
What does the Article 13 appropriateness guideline ask of companies?
The Headquarters decided the Guideline for Ensuring the Appropriateness of Research & Development and Utilization of Artificial Intelligence-Related Technology on December 19, 2025 (English text: provisional translation). It says it aims “to encourage voluntary and proactive efforts by all stakeholders,” and that stakeholders “are required to respond appropriately at a suitable level, taking into account their scale, position, and the risks posed by AI, as well as the technologies and knowledge available at the time.” The Japanese text says 対応することが求められる (our translation: “are called on to respond”), the document aims at voluntary efforts, and no sanction attaches to it. For businesses that develop and provide AI, it lists five matters:
- AI governance across the lifecycle, including processes to identify, assess and address risks, with monitoring involving management.
- Explainability, within reasonable limits, about the origin of training data and outputs, and information for users such as limitations, prohibited actions, training-data collection policies and warnings on output reliability.
- Measures against misuse for crime; developers and operators “will strive to develop technologies that can determine whether something is generated by AI” (such as digital watermarks, provenance management and APIs) and implement them as necessary.
- A business continuity plan, set in advance.
- Communication with data holders, including rights holders; businesses whose AI has significant social impact endeavor to consider returning benefits to them.
What are the AI Guidelines for Business?
These guidelines do not bind: they call themselves “non-binding soft law.” The Ministry of Internal Affairs and Communications (MIC) and the Ministry of Economy, Trade and Industry (METI) publish them: version 1.0 on April 19, 2024, 1.01 on November 22, 2024, 1.1 on March 28, 2025, and 1.2 on March 31, 2026. They set ten common guiding principles, from human-centricity and safety to fair competition and innovation, and separate tasks for AI developers, AI providers and AI business users. The English versions are provisional translations, and Appendices 7 to 9, including the checklist, are in Japanese only. Developers who develop advanced AI systems in compliance with the Hiroshima Code of Conduct are “expected to participate” in its reporting framework.
How does Japan’s privacy law apply to AI?
The Act on the Protection of Personal Information (Act No. 57 of 2003) applies to AI training and use like any other processing. Under Article 171, it also applies when a business handles, in a foreign country, personal information by which a person in Japan is identifiable, in relation to supplying a good or service to that person.
On June 2, 2023, the Personal Information Protection Commission (PPC) issued an alert on generative AI services (Japanese only). It told businesses entering personal information in prompts to confirm that this is within the scope necessary to achieve the specified purpose of use, and warned: 「あらかじめ本人の同意を得ることなく生成AIサービスに個人データを含むプロンプトを入力し、当該個人データが当該プロンプトに対する応答結果の出力以外の目的で取り扱われる場合、当該個人情報取扱事業者は個人情報保護法の規定に違反することとなる可能性がある」 (our translation: “if a business enters prompts containing personal data into a generative AI service without the person’s prior consent and the data is handled for purposes other than outputting the response, the business may violate the APPI”). The alert therefore tells businesses to confirm that the provider does not use the data for machine learning. A separate June 1, 2023 alert to OpenAI under APPI Article 147 addressed acquiring special care-required personal information without prior consent, except in the cases listed in Article 20(2), and notifying or publishing purposes of use in Japanese.
The 2026 amendment passed the Diet on July 10, 2026 and was promulgated on July 17, 2026. Among other changes, the PPC’s outline (Japanese only) says it will:
- Make consent unnecessary for providing personal data to third parties, and for acquiring special care-required personal information that has been made public, where the data is used only to prepare statistical information, including AI development that can be classified as statistical compilation (new Articles 30-2 and 31-3). Conditions include publishing certain matters and, for provision, a written agreement that it is only for preparing statistical information.
- Require notice of certain matters about handling facial feature data and similar data, ease the requirements for requests to stop using it, and bar its provision under the opt-out system.
- Let the PPC order a surcharge equal to the financial benefit gained, in cases such as where a serious violation has infringed individuals’ rights and interests (new Articles 148-3 to 148-17), limited to listed violation types, to businesses that failed to exercise due care (相当の注意) to prevent the violation, to cases where individuals’ rights and interests were infringed or a concrete risk of infringement arose, and to large-scale cases, with 1,000 affected persons as the benchmark.
In principle it takes effect on a date set by Cabinet Order within two years of promulgation. On November 12, 2026, the PPC’s page showed the implementing rules in preparation and gave no effective date.
Can AI be trained on copyrighted works in Japan?
Often, under Article 30-4 of the Copyright Act (Act No. 48 of 1970). It permits exploiting a work “in any way and to the extent considered necessary” in listed cases, including data analysis, “or in any other case in which it is not a person’s purpose to personally enjoy or cause another person to enjoy the thoughts or sentiments expressed in that work; provided, however, that this does not apply if the action would unreasonably prejudice the interests of the copyright owner in light of the nature or purpose of the work or the circumstances of its exploitation.”
The Legal Subcommittee under the Copyright Subdivision of the Cultural Council issued the General Understanding on AI and Copyright on March 15, 2024 (English overview, May 2024). It is “not legally binding.” Its main points:
- Article 30-4 does not apply where a purpose of enjoyment coexists, such as fine-tuning meant to generate material similar to training works, or retrieval-augmented generation (RAG) that aims to output their creative expression; Article 47-5 may still permit some RAG uses.
- The proviso applies where a database work available on the internet for a fee, in a format usable for information analysis, is reproduced without compensation. The overview says reproducing such database works while avoiding technical measures against AI-training reproduction, such as robots.txt instructions, falls within the proviso.
- Output infringes where it has similarity to and dependence on an existing work and no exception applies. If the work was used to train the AI, dependence is generally assumed even if the user did not know.
What are the AI Safety Institute and the Hiroshima AI Process reporting framework?
The Japan AI Safety Institute (AISI) was launched on February 14, 2024, with its secretariat at the Information-technology Promotion Agency (IPA). It works on AI safety evaluation methods and standards; its 2025 guides cover red teaming, evaluation perspectives and data quality. The AI Promotion Act does not mention AISI and gives it no enforcement role. On May 29, 2026, AISI and OpenAI signed a memorandum of cooperation on AI safety evaluation.
The Hiroshima AI Process Comprehensive Policy Framework, including the International Code of Conduct for Organizations Developing Advanced AI Systems, was agreed by G7 digital ministers and endorsed by G7 leaders in December 2023. Its voluntary Reporting Framework launched on February 7, 2025. MIC reports that 19 companies’ first responses were published on the OECD website in April 2025 and that responses to Reporting Framework 2.0 from 31 companies were published on October 2, 2026.
Which Japanese AI rules bind companies?
| Rule | Binding? | Who it reaches | Status and date |
|---|---|---|---|
| AI Promotion Act (Act No. 53 of 2025) | Yes, but no penalty provisions; duty to cooperate (Art. 7) | Businesses that develop, provide or use AI in their business; the Article 13 guideline reads this to include overseas operators | Partly in force June 4, 2025; fully in force September 1, 2025 |
| AI Basic Plan (Art. 18) | No; a Cabinet plan for government measures | Government | First plan December 23, 2025; second plan July 14, 2026 |
| Guideline for Ensuring the Appropriateness of AI R&D and Utilization (Art. 13) | No; aims at voluntary efforts | R&D institutes, utilization business operators, governments, citizens | Decided by the Headquarters December 19, 2025 |
| AI Guidelines for Business | No; “non-binding soft law” | AI developers, AI providers, AI business users | Version 1.2, March 31, 2026 |
| APPI (Act No. 57 of 2003) | Yes; PPC supervision | Businesses handling personal information, including abroad in relation to supplying goods or services to a person in Japan (Art. 171) | In force |
| APPI amendment (Act No. 56 of 2026) | Yes, once in force | Businesses handling personal information | Promulgated July 17, 2026; in principle in force on a date set by Cabinet Order within two years |
| PPC alert on generative AI services | An alert; it states when use may violate the APPI | Businesses and administrative bodies using generative AI; separate alert to OpenAI | June 2, 2023 |
| Copyright Act Article 30-4 | Yes | Anyone exploiting works, including for AI training | In force; current e-Gov text checked November 12, 2026 |
| General Understanding on AI and Copyright | No; “not legally binding” | AI developers, service providers and users | March 15, 2024; English overview May 2024 |
| Hiroshima AI Process Reporting Framework | No; voluntary | Organizations developing advanced AI systems | Launched February 7, 2025; 2.0 responses published October 2, 2026 |
How does Japan compare with South Korea, China and the EU?
Japan’s Act sets no fines, risk tiers, labeling duty or local-representative duty. South Korea’s AI Basic Act, in force since January 22, 2026, sets transparency, high-impact and domestic representative duties, with administrative fines capped at 30 million won; the Ministry of Science and ICT has announced a grace period of at least one year deferring fact-finding investigations and penalties, except in highly exceptional cases. As of November 3, 2026, China had no comprehensive AI law and regulated by service type through filing and labeling rules. The EU AI Act prohibits listed practices and sets duties for high-risk systems. In Japan, binding exposure comes from the APPI and the Copyright Act.
What should a foreign company selling AI into Japan do?
The first four items concern binding law. The rest are preparation.
- Map personal information of people in Japan that you handle abroad in relation to supplying them goods or services (APPI Article 171)
- Confirm your AI vendor does not use prompts containing personal data for machine learning, or obtain prior consent
- Wait for the PPC’s rules and effective date before relying on the Act No. 56 of 2026 statistical-purpose exception
- Check training-data collection and RAG uses against Article 30-4, its proviso and the General Understanding
- Treat the business as an Article 7 utilization business operator and name a contact for government inquiries
- Map your practices to the voluntary Article 13 guideline and AI Guidelines for Business, version 1.2, including the checklist in the Japanese-only appendices
- If you develop advanced AI systems, decide whether to submit a Hiroshima AI Process Reporting Framework response
- Consider a complaint-handling system; the second AI Basic Plan commits the government to measures enabling businesses to establish their own but places no duty on them
How this fits the other guides: compare the South Korea AI Basic Act guide, the China AI regulations guide and the EU AI Act compliance guide; the AI compliance guide places cross-border AI laws in its first layer.
This guide is general information as of November 12, 2026, not legal advice. Confirm the current Japanese texts and the government’s position with counsel admitted in Japan or with the Cabinet Office, the PPC or the Agency for Cultural Affairs before acting.
Common questions
Does Japan have an AI law?
Yes. The Act on Promotion of Research and Development, and Utilization of Artificial Intelligence-related Technology (Act No. 53 of 2025) took full effect on September 1, 2025. It sets basic principles, an AI Basic Plan and an AI Strategic Headquarters, and contains no penalty provisions; binding rules for AI also come from existing law such as the APPI and the Copyright Act.
What does Japan's AI Promotion Act require of companies?
Under Article 7, a business that intends to develop or provide AI products or services, or to use AI in its business activities, must cooperate with the measures national and local governments implement under Articles 4 and 5, in accordance with the basic principles. Under Article 16, the government analyzes cases in which people's rights and interests are infringed through AI used for wrongful purposes or by inappropriate methods, and gives guidance, advice and information or takes other necessary measures. The Act sets no fines and no sanction for failing to cooperate.
Can AI be trained on copyrighted works in Japan?
Often. Article 30-4 of the Copyright Act permits exploiting a work in any way and to the extent considered necessary, for data analysis and other cases in which it is not a person's purpose to personally enjoy or cause another person to enjoy the thoughts or sentiments expressed in it, unless the action would unreasonably prejudice the copyright owner's interests in light of the nature or purpose of the work or the circumstances of its exploitation. The non-binding General Understanding on AI and Copyright issued by the Legal Subcommittee under the Copyright Subdivision of the Cultural Council on March 15, 2024 says the exception does not apply where a purpose of enjoyment coexists.
By industry: all 8 industry pages
Latest: Japan AI policy news, each entry dated and sourced.
Sources
- Act on Promotion of Research and Development, and Utilization of Artificial Intelligence-related Technology (Act No. 53 of 2025), Japanese/English, Ministry of Justice, Japanese Law Translation, Act of June 4, 2025.
- Japanese Law Translation: search page and notes on translations, Ministry of Justice, accessed November 12, 2026.
- 人工知能関連技術の研究開発及び活用の推進に関する法律 (current Japanese text), e-Gov Law Search, Digital Agency, revision in force September 1, 2025.
- 人工知能関連技術の研究開発及び活用の推進に関する法律(AI法) (in Japanese), Cabinet Office.
- Outline of the Act on Promotion of Research and Development, and Utilization of AI-related Technology (Provisional Translation), Cabinet Office; Japanese version: AI法の概要.
- 人工知能戦略本部の設置等について (in Japanese), Cabinet Office, September 1, 2025.
- 城内内閣府特命担当大臣記者会見要旨 (minister’s press conference, in Japanese), Cabinet Office, February 28, 2025.
- 小野田内閣府特命担当大臣記者会見要旨 (minister’s press conference, in Japanese), Cabinet Office, January 16, 2026.
- 人工知能基本計画 (AI Basic Plan page, in Japanese with English links), Cabinet Office.
- Japan’s Second Artificial Intelligence Basic Plan (Provisional translation), Cabinet Office, Cabinet decision of July 14, 2026; Summary.
- Artificial Intelligence Basic Plan (Provisional translation), Cabinet Office, Cabinet decision of December 23, 2025.
- 人工知能関連技術の研究開発及び活用の適正性確保に関する指針 (guideline page, in Japanese), Cabinet Office, decided December 19, 2025.
- Guideline for Ensuring the Appropriateness of Research & Development and Utilization of Artificial Intelligence-Related Technology (Provisional translation), AI Strategic Headquarters, December 19, 2025.
- 「AI事業者ガイドライン」掲載ページ (AI Guidelines for Business page, in Japanese with English links), Ministry of Internal Affairs and Communications.
- AI Guidelines for Business Ver1.2 (Provisional Translation), Ministry of Internal Affairs and Communications and Ministry of Economy, Trade and Industry, March 31, 2026; Outline of AI Guidelines for Business Ver1.2, which says Appendices 7, 8 and 9 are Japanese only.
- Act on the Protection of Personal Information (Act No. 57 of 2003), English, Ministry of Justice, Japanese Law Translation.
- 生成AIサービスの利用に関する注意喚起等について (in Japanese), Personal Information Protection Commission, June 2, 2023; PDF.
- 令和8年 改正個人情報保護法について (in Japanese), Personal Information Protection Commission; 個人情報保護法等の一部を改正する法律について (outline, July 2026).
- Copyright Act (Act No. 48 of 1970), Japanese/English, Ministry of Justice, Japanese Law Translation.
- “General Understanding on AI and Copyright in Japan”: Overview, Agency for Cultural Affairs, May 2024; Japanese original: AIと著作権に関する考え方について, March 15, 2024.
- AISIについて (About AISI, in Japanese), Japan AI Safety Institute.
- 米国OpenAI社と日本AISIのMOC締結について (in Japanese), Cabinet Office, May 29, 2026.
- About the Hiroshima AI Process, Ministry of Internal Affairs and Communications; Reporting Framework; Documents of Achievement.