Japan Drafts First AI Basic Plan, Putting Public Institutions at the Front of Adoption

Government services are at the center of Japan's AI push

Japan has drafted its first national AI Basic Plan, with Cabinet approval targeted before the end of 2025, to make the country the most AI friendly place for development and use. Ministries, agencies and municipalities would lead adoption, using artificial intelligence to improve efficiency and public services. Defense applications are also proposed.

As of September 20, 2025, the plan remains a draft. It is due to go before the AI Strategy Headquarters, chaired by Prime Minister Shigeru Ishiba, before an expert panel refines its details. The headquarters brings together all Cabinet ministers, placing AI policy under the prime minister's direction.

The push responds to modest domestic adoption. Surveys indicate that about 20 percent of individuals and roughly half of companies used AI last year. A separate measure found that 26.7 percent of people used generative AI in fiscal 2024, well below rates in the United States and China. Generative AI produces material such as text or images in response to instructions.

Japan's plan combines faster adoption with management of risks, including incorrect answers, disinformation, privacy harms and national security threats. Its central idea is collaboration between people and AI: software can handle demanding information tasks, while staff remain responsible for decisions and oversight.

The draft rests on an AI law enacted in May 2025. That framework establishes national institutions and responsibilities for AI development and use, while favoring guidance and transparency over punitive fines. The government intends to investigate serious infringements of rights involving AI and help shape international rules for safe, trustworthy technology.

Four policies connect adoption, development and oversight

The draft sets out four basic policies: accelerate AI use, strengthen development capacity, lead in AI governance, and sustain innovation toward a society in which AI is widely integrated. Together, they address both the tools Japan needs and the institutions responsible for their use.

Public bodies would lead the first policy by streamlining routine work and improving services. The second focuses on the foundations needed to develop capable systems: quality data, computing resources and skilled people. The third connects domestic guidance, investigations and international principles. The fourth calls for education, workforce training and institutional changes that support continued collaboration between people and AI.

Copyright and civil liability run through the plan. Copyright concerns who may use protected material, including in training data and generated outputs. Civil liability concerns who bears legal responsibility when harm occurs. Clearer rules would help public institutions and businesses understand their obligations before deploying systems.

The plan also identifies quality data as a Japanese strength. Manufacturing, healthcare and transport generate structured information that could improve AI accuracy when it is properly prepared and protected. Development capacity therefore means more than building models: it includes making information usable, securing it and training people to judge the results.

What AI could change in public services

Ministries and municipalities could begin with document processing, automated translation and summarization. AI assistants could support call centers and pass complex problems to human staff. Systems could also sort benefits and permit cases for attention, helping officials manage queues and reduce delays.

Local governments could use AI for disaster preparedness, hazard mapping, evacuation planning and information during storms and earthquakes. Health services could prioritize screenings, flag possible adverse events and improve scheduling. Transport authorities could adjust traffic signals and fleet dispatch, while courts and regulators could search case law and identify patterns in filings.

Japan's aging population and labor shortages give these proposals practical urgency. Smaller teams need to handle substantial workloads with consistent quality. AI could compare policy scenarios quickly or scan large datasets for possible fraud, tax gaps and safety violations.

Human staff would remain accountable. An AI system might identify an unusual record or suggest a first draft, but an appropriate official or specialist would review the work. This division of tasks matters particularly where public decisions affect benefits, health or safety.

The intended gains are concrete: fewer backlogs, less routine work and quicker answers for residents. Achieving them would require agencies to evaluate accuracy and service quality, alongside the time saved.

Data, computing power and skills need investment

Japan's manufacturing, health and transport data could support more accurate models, provided it is curated carefully. The draft highlights reliable data pipelines, consistent labeling and privacy safeguards. A data pipeline is the process that moves information from collection through preparation to use; labels explain what records contain so a model can learn from them.

More open data, where feasible, would give universities, startups and established companies material for building and evaluating models. Computing power, storage and networking also need investment. Models adapted to Japanese language and context are another priority, because language and specialist terminology can affect how useful a system is.

Civil servants, engineers and specialists in individual fields would need education and reskilling to use AI responsibly. Staff must be able to assess outputs, recognize limitations and understand when a task requires human judgment.

The draft also points toward procurement changes that would let agencies run pilots, assess performance and specify security, privacy and quality requirements. A broader range of suppliers, including small companies and research consortia, could support experimentation while maintaining safeguards.

Existing practices make adoption harder. Paper processes and older computer systems remain common, while datasets can be isolated in separate departments or stored in incompatible formats. Procurement cycles often favor large projects lasting several years over small trials that expand after demonstrating results. Organizations may also respond to compliance concerns by avoiding AI instead of managing its risks.

Clear guidance, shared tools and institutional support are intended to help agencies and businesses address those barriers. Private AI investment in Japan also trails other major economies by a wide margin, adding to concern that domestic companies could fall behind competitors already using AI in products, operations and customer service.

How the AI law changes government oversight

The AI law enacted in May 2025 defines policy goals, institutions and responsibilities for research, development and use of AI technologies. It establishes an AI Strategy Headquarters in the Cabinet and an AI Strategy Center to guide the Fundamental AI Plan, develop guidelines and coordinate ministries.

The framework covers developers, providers and business users, including foreign operators supplying AI used in Japan. Its core principles include productivity, transparency, privacy, fairness and safety. Businesses are expected to make reasonable efforts to follow government principles and cooperate with investigations.

Authorities can survey practices, issue guidance and investigate serious cases involving infringements of rights and interests. The approach does not impose broad penalties comparable to those in the European Union, but gives the government a route to request information, provide instructions and disclose findings in significant cases.

Chief Cabinet Secretary Yoshimasa Hayashi described the government's balance between the benefits of AI and the risks of misuse:

AI offers benefits like improved productivity and solutions to labor shortages, it also poses risks such as disinformation and sophisticated criminal activity.

Agencies would survey AI use and publish guidance. Businesses and research institutions would be expected to document the stages of an AI system's development and use, manage data rights, and address bias, safety and security from design through deployment.

For an organization adopting AI, this means keeping track of more than the finished product. Data collection, model training, deployment and later monitoring all matter, because problems can arise at any of those stages. The Basic Plan calls for risk management throughout that process.

Privacy, copyright and responsibility remain central

The Act on the Protection of Personal Information, known as the APPI, remains central to privacy compliance. It governs consent, data minimization, breach reporting and transfers across borders. Data minimization means limiting personal information to what is needed for the task.

Agencies and vendors would need to justify their use of data and maintain strong access controls. Techniques such as pseudonymization, which replaces identifying details with substitute identifiers, form part of the privacy measures described in the plan.

Copyright questions involving training data and AI outputs remain under review. The draft seeks to protect intellectual property while enabling fair use and innovation, with particular relevance to media, software and design. It also calls for clearer civil liability frameworks so organizations understand responsibility when AI use causes harm.

Disinformation and harmful content are another concern. The framework connects these issues with faster takedown mechanisms under recent legislation for information distribution providers. AI governance would therefore operate alongside existing privacy, copyright and information rules.

Transparency is especially important when systems affect benefits, healthcare or public safety. Agencies would need to explain when and why AI is being used, the expected benefits and risks, and how people can seek redress if something goes wrong. Human support and public feedback would help institutions identify problems after deployment.

The plan aligns domestic guidance with international principles, including the Hiroshima International Guiding Principles for safe, trustworthy AI centered on people. Japan intends to use that work to help shape international approaches to safety evaluation, copyright, disinformation response and transparency.

Defense uses bring security demands

The draft proposes adopting AI to strengthen defense capabilities and protect critical infrastructure. Possible uses include logistics planning, predicting maintenance needs, cyber defense, analysis of satellite and sensor information, and support for decisions in complex operations.

Defense personnel could use these tools to process more information and react faster. The same technology also creates threats, including deepfake propaganda, advanced cyberattacks and autonomous threats. Deepfakes are fabricated or altered media that can make people appear to say or do things they did not.

Prime Minister Shigeru Ishiba stressed the security stakes at the headquarters meeting on AI strategy:

AI is very crucial in terms of security,

Ishiba also called for speed as competition in global AI development intensifies, saying Japan must:

implement support measures promptly

Defense adoption would require rigorous reliability checks, testing against attempts to manipulate systems and careful human oversight. The draft's focus on quality data and disciplined governance applies directly to those requirements. International cooperation on safe AI practices and responsible military use is also intended to reduce the risk of escalation.

Who will develop the plan and put it into practice?

The AI Strategy Headquarters, established in the Cabinet Office and chaired by the prime minister, is steering the process. All Cabinet ministers participate. An expert panel of researchers, industry leaders and policy specialists would refine the draft before the government seeks Cabinet approval later in 2025.

The AI Strategy Center would coordinate implementation across ministries and agencies and continue publishing guidelines and surveys tracking practice. Several other bodies would retain responsibilities within their existing fields.

The Ministry of Economy, Trade and Industry and the Ministry of Internal Affairs and Communications jointly prepare business guidelines. The Agency for Cultural Affairs guides copyright matters, while the Personal Information Protection Commission guides privacy. Regulators in fields such as finance and health would align sector guidance with the wider plan.

The Digital Agency and local governments would be central to deployment because many public services involve direct contact with residents. National coordination would therefore need to connect with decisions about local systems, staffing and service delivery.

The policy sequence spans several stages:

• May 2025: Japan enacted the AI law establishing the framework for national AI policy and its coordinating institutions.
• September 20, 2025: The first national AI Basic Plan remains a draft, with presentation to the AI Strategy Headquarters and expert refinement ahead.
• Later in 2025: The government aims to secure Cabinet approval. No exact approval date has been given.

What agencies can prepare before approval

Public institutions do not need to wait for final approval to prepare. The draft points toward narrowly defined projects where useful data, measurable goals and clear oversight are already available. Agencies could assess whether a trial reduces document turnaround times or improves the share of contact center inquiries resolved.

Governance would need to be part of a project from the beginning. That includes plans for audits, documentation, communication with users and staff training. Expansion would follow evaluation, rather than being assumed when a pilot begins.

Practical preparations include:

• Assign a senior official to coordinate AI work and align projects with privacy, security and accessibility requirements.
• Map available data, classify its sensitivity, and establish quality, labeling and access procedures consistent with the APPI.
• Start with narrow pilots and measurable service goals, then assess results before expanding.
• Set vendor contract terms covering data rights, copyright, liability, incident reporting and model quality assurance.
• Tell users when and how AI is used, provide access to human support, and record feedback to improve services.

Agencies would also need staff who can challenge an incorrect output. Faster processing is useful only when the service remains dependable, particularly where a resident relies on an answer about a benefit, permit or health matter. The draft's collaboration model keeps that responsibility with people.

Continuous innovation and Japan's international role

The fourth policy extends beyond initial deployment. Education, reskilling and partnerships between universities, startups and established companies would support continued development. Laws, benefits systems and workforce policies would also need regular updates as AI advances.

AI combined with robotics and sensors, sometimes called physical AI, is part of this approach. These systems perceive and act in the physical world. In eldercare, logistics and inspection, they could move routine tasks from scarce personnel to machines supervised by people.

Public communication would be central to trust. Institutions would need to explain their decisions about AI use, listen to feedback and provide routes for addressing harm. Continued innovation therefore involves changing working practices and oversight alongside technology.

Japan is also positioning its approach among different international models. The European Union is moving toward a comprehensive legal regime for systems considered high risk. The United States relies more on agency guidance and executive action. Japan combines a national framework with voluntary guidance and sector rules, while reserving investigations and public disclosure for serious cases.

The Hiroshima initiative gives Japan a basis for contributing to international principles. Its strengths in manufacturing, robotics and precise data collection could also support practical standards for trustworthy systems.

The domestic task is to connect those strengths with usable data, workable procurement and consistent governance. The plan seeks to help public institutions and companies improve productivity while keeping responsibility for decisions visible and giving people a way to challenge problems.

The Bottom Line

• Japan's first AI Basic Plan remains a draft as of September 20, 2025, with Cabinet approval targeted later in the year.
• Its four policies cover faster adoption, development capacity, governance and continuous innovation.
• Public institutions would lead adoption, with defense uses also proposed.
• Surveys show about 20 percent of individuals and roughly half of companies used AI last year; generative AI use reached 26.7 percent in fiscal 2024.
• The May 2025 AI law establishes coordinating institutions and powers to investigate serious infringements of rights.
• Quality data, computing resources, staff training and procurement changes are central to implementation.
• Privacy, copyright, civil liability and human oversight remain key requirements as agencies prepare practical trials.