Government AI Platform 'Gennai' Goes Open Source | How to Win Public Sector AI Projects

What Is "Gennai (GENAI)"? — An Overview of the Government AI Platform Released by the Digital Agency

Gennai (GENAI) is the unified government AI platform developed by Japan's Digital Agency. It is designed as a system that allows approximately 180,000 ministry and agency staff to access AI functions through a common interface, with the goal of centralizing the cost and administrative burden of each ministry adopting its own disparate AI tools (Impress Watch report).

The most notable aspect of this OSS release is that "elimination of vendor lock-in" is explicitly stated as a design principle. According to the Digital Agency's large-scale deployment materials, the system is structured to avoid dependence on any specific cloud provider or model vendor, prioritizing procurement flexibility and competitive principles. The lesson learned from government systems being locked into specific vendors led to the choice of OSS.

Two main types of components have been released. The first is the Web interface — the frontend and API layer through which staff invoke AI functions from a browser. The second is infrastructure deployment templates organized for three cloud providers: AWS, Azure, and GCP. Note that foundation models such as ChatGPT or Claude are not included. Model weights and government-internal data are not part of the public release; users must separately procure and connect a model of their choosing.

Large-scale proof-of-concept testing is planned for fiscal year 2026, which is expected to include RAG (Retrieval-Augmented Generation) templates and legal/regulatory AI implementation (innovatopia explainer). The templates are expected to grow richer based on the proof-of-concept results, so there is ample value in getting familiar with them now (※ for the latest status of the proof-of-concept, please check the Digital Agency official site).

What Can You Use Under the MIT License? — Contents and Scope of the Open-Source Code

The MIT License is one of the most permissive open-source licenses available. Its core terms are as follows: "Permission is hereby granted, free of charge, to any person obtaining a copy of this software ... to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software" — provided that the copyright notice and license terms are included. In short, commercial use, modification, and redistribution are all permitted, and it is perfectly legal to incorporate the code directly into a system delivered to a municipal client.

Specific usable components include, first, the Web interface: you can reference and reuse the staff-facing chat UI and API gateway implementation as the skeleton for systems with similar requirements for other municipalities. Second are the cloud templates: configuration definitions for AWS, Azure, and GCP are included, enabling significant reduction of infrastructure build time.

On the other hand, it is equally important to be clear about what is not available. Foundation model weights for GPT-4, Claude, Gemini, and similar models are not included in Gennai. Nor are classified data accumulated within the government or the portions implementing individual ministries' proprietary business rules, which are naturally not disclosed. Municipality-specific ordinance and regulation data must also be prepared independently.

Regarding deployment to private companies and municipalities: while there are no issues from a licensing standpoint, compliance with security requirements and government information system standards remains a separate matter. The excessive optimism that "MIT means total freedom" is a mistake — a point we will address in more detail in the section on entry conditions below.

Reading the Current Landscape of Public Sector AI Projects — Market Size, Procurement Flow, and Entry Conditions

Government digitalization investment, including supplementary budgets for fiscal year 2026, continues to grow. As can be confirmed on the Digital Agency's Japan Dashboard, digitalization-related expenditures in local government finances have increased year over year, and the market for municipal AI adoption is estimated to be in the hundreds of billions of yen (※ based on estimates from multiple industry research reports). Gennai's open-sourcing can be understood as creating an environment in which freelancers and small engineers can now seriously enter this market.

Let us organize the procurement flow. At the top are national-level projects procured by the Digital Agency and the Ministry of Internal Affairs and Communications as policy initiatives — these typically go to large SIers or consulting firms as prime contractors. Next are IT procurements independently issued by prefectures and designated cities, traditionally another domain where large players are strong. The most accessible entry point is municipal-level procurement: because budget scales are smaller, mid-sized and freelance engineers are more likely to be approached, and using Gennai's reference implementation can strengthen the persuasiveness of your proposals.

There are three entry conditions you need to realistically understand. First, security requirements: you need to verify on a project-by-project basis whether NISC standards and ISMS certification requirements apply. Second, organizational requirements: you may be asked to establish a PMBOK-compliant project management structure and an incident response contact point. Third, track record requirements: you may be asked to demonstrate a delivery record in similar projects.

For finding projects that freelancers and small firms can enter, regularly checking e-Gov and JKKN (the electronic bidding core system used by municipalities nationwide) is the baseline. Searching bid announcements with keywords such as "AI," "chatbot," "conversational," and "operational efficiency" makes it easier to find projects with high affinity to Gennai. Participating as a subcontractor to a prime contractor is also a valid path, and actively registering as a partner with large SIers is a worthwhile step.

Implementation Approach to Win Contracts Using Gennai — RAG and Legal/Regulatory AI Implementation

Here is the basic procedure for customizing the RAG template for municipal use. The Gennai RAG template stores documents in a vector database and, in response to a user's question, retrieves relevant documents before the AI generates its answer. When customizing for a municipality, the starting point is swapping the target documents for "that municipality's ordinances, regulations, guidelines, and FAQs." You should expect significant effort in document preprocessing (PDF text extraction, table parsing, header/footer removal).

The biggest caution when implementing legal/regulatory AI is handling the amendment cycles of ordinances and regulations. Ordinances can be amended at each regular session of the local assembly, and unless version management for documents and the update frequency of the RAG index are built in from the design stage, there is a risk of generating incorrect answers based on outdated text. From the standpoint of reliability required of administrative systems, it is also important to design the UI to explicitly state "the answer is based on the text as of [date]."

Regarding which of the three cloud providers to choose: in many cases, the municipality's existing cloud contracts are the biggest determining factor. For a municipality already running its basic resident register system on Azure, choosing the Azure template is the natural choice. Considering data transfer costs and network configurations between clouds, proposing the option with the lowest integration cost with existing infrastructure is rational from the perspective of improving your chances of winning the contract.

The quickest steps to build a "working demo" to show in the proposal phase are as follows. ① Launch Gennai's Web interface locally, ② scrape text from the municipality's publicly available FAQ page and load it into the vector database, ③ demo with questions representative of inquiries municipal staff handle daily — with these 3 steps, you can prepare a prototype within a few days. More than polish, the fact that it is "running on your agency's actual data" dramatically increases the persuasiveness of your proposal.

3 Steps to Take Action Right Now — From Repository Setup to Proposal Writing

Step 1: Get Gennai Running Locally

Access the Gennai repository from the Digital Agency's official GitHub account (https://github.com/digital-go-jp) and set up your environment following the README setup instructions. The Web interface is set up so that launching it only requires Docker and can be completed in a short time. First, confirm it runs locally and understand the roles and dependencies of the components. For the cloud templates, we recommend starting by reading the one for the cloud provider you are most familiar with. Actually working through the code will make it clear "where the municipality-specific customization points are."

Step 2: Regularly Monitor Procurement Information Sources

The primary sources of procurement information you should monitor are as follows. e-Gov (https://www.e-gov.go.jp) aggregates national bid information and serves as the starting point for finding ministry and agency projects. The integrated e-government portal also covers national procurement information. For prefectures and municipalities, check each local government's "Bids & Procurement" page directly, or if the municipality uses JKKN's electronic bidding system, a consolidated search is possible. Setting up RSS feeds or email notifications to reduce the daily monitoring cost is recommended.

Step 3: Include Gennai OSS-Based Implementation Rationale in Your Proposal

The greatest benefit of using Gennai in your proposal is the credibility of being "based on an implementation designed and released by the government itself." Stating explicitly in the technical approach section of your proposal that you are "adopting the MIT-licensed government AI platform (Gennai) published by the Digital Agency as a reference implementation" makes it easier to explain specification transparency and maintainability compared to a fully custom build. Also, given Gennai's nature as OSS, the low future vendor lock-in risk is a strong selling point for procurement officers.

The Right Time to Act: In Line with the Fiscal Year 2026 Large-Scale Proof of Concept

The Digital Agency plans to conduct a large-scale proof of concept for Gennai during fiscal year 2026. As the RAG templates and legal/regulatory AI implementation become more robust, a wave of municipalities considering "can we use this too?" is sure to arrive. To ride that wave, it is important to start experimenting with implementations now and build connections with municipal staff. The proof-of-concept results will be published in due course, but engineers who are already moving will be in a more advantageous position when project volume increases after the proof of concept.

• Gennai OSS Release – Impress Watch
• Gennai Large-Scale Deployment Materials – Digital Agency
• Gennai OSS Design Philosophy Explained – innovatopia
• Japan Dashboard Local Government Finance – Digital Agency

Sources

• Anthropic Official News
• Anthropic Engineering Blog
• Claude Official Documentation