Legal Brain
: How to use it, features, and the business problems it solves

What is Legal Brain

Legal Brain is the name of an AI foundation technology specialized for the legal domain, developed by Bengo4.com, Inc. The official notation is LegalBrain (one word), and the product built on it is the "LegalBrain Agent." The agent launched on May 23, 2025, with legal research as its first function.

The underlying technology, Legal Brain 1.0, is officially described as "an AI foundation technology that comprehensively learns legal data such as statutes, court precedents, guidelines, and specialist books, achieving advanced language understanding specialized for Japanese legal context." The starting point is that it is not a general-purpose large language model tuned for legal work, but something built for the Japanese legal context.

Source: Bengo4.com, Inc. press releases (the diagrams below are from the same source)

Attacking the assumption that "AI gets things wrong" through citations

What kept AI out of legal work was not accuracy so much as how you verify an answer. When article numbers and case citations come back in a plausible-looking form, you end up researching from scratch to confirm them.

Legal Brain is described as generating answers based only on trustworthy sources, thereby suppressing to the greatest extent possible the hallucinations that were a problem with general-purpose LLMs. The footnote markers at the end of each statement in the screen above, and the reference list in the right pane showing title, author, and publication date, reflect that design. The "research scope 148" indicator shows how many documents the answer was drawn from.

The background the company cites is survey data (November 2024) showing that 54.5% of legal staff name "understaffing" and 52.7% name "sheer volume of work" as their challenges. Research is high in volume yet specialized enough that it is hard to outsource.

The data Bengo4.com has accumulated, turned into a graph

At the core of Legal Brain is a proprietary database called LegalGraph.

What happens between the question and the answer

The official architecture diagram makes clear that this is not the kind of product that simply throws a question at an AI and has it write an answer.

The flow works as follows.

1. The user enters a question in natural language
2. That input is sent to Legal Brain
3. Legal Graph and the search engine exchange data through nearest-neighbor search, narrowing down related documents
4. The narrowed set of documents is sent to the AI
5. The AI returns bullet-point notes and passages from multiple related documents
6. The result is displayed in the agent

Information is extracted in advance from the document database (books, statutes, precedents, guidelines) and handed to the search engine. In other words, the AI does not compose text freely — the structure identifies the documents first, then has the model organize based on them. That ordering is what makes citations possible.

Legal Brain is not the name of a single product

One more thing worth knowing: Legal Brain is not a standalone product but a foundation technology supporting several Bengo4.com services. The official site lists the following as the service lineup built around Legal Brain.

An integrated AI legal research tool powered by Legal Brain had already launched on August 30, 2024, with two functions: "extraction and organization of legal issues" and "information gathering." The 2025 agent is an extension of that line.

How to use

1. Apply for access

   Make an inquiry or apply through the official site. It targets attorneys, law firms, and corporate legal departments, and contracts are made at an organizational level.

2. Enter your question in natural language

   Type the issue as prose. There is no need to reformulate it into keywords or learn search operators. In the screen shown above, the sentence "Can liability for damages based on product liability be directly excluded by contract?" is entered as-is.

3. Review the organized legal issues

   The AI extracts the legal issues and presents them structured as bullet points. In the example above, it breaks the question into "whether liability can be limited toward consumers" and "the permissible scope of liability limitation between businesses."

4. Go to the primary sources through the references

   Each statement carries a footnote, and the right pane lists references with title, author, publisher, and publication date. Checking the primary source here is part of a single unit of work.

5. Share the results

   Research results can be saved and shared within your team, accumulating as knowledge per matter.

Features

Pricing

Pricing is not published; it is quoted individually. No amounts appear on the official site, and figures are presented through an inquiry.

As of August 2026, the official site does not state amounts or whether a free trial is available. Please confirm directly with the provider when considering adoption.

Adoption and reception

Publicly disclosed adopters include Anderson Mori & Tomotsune, Mizuho Bank, and Mizuho Securities — a combination of one of Japan's leading law firms and megabank-affiliated legal departments, showing that adoption has begun in areas with the least tolerance for incorrect answers.

According to the November 11, 2025 announcement, Anderson Mori & Tomotsune verified the tool's usefulness in a trial and then began full-scale use across all attorneys and paralegals. What the trial rated highest was "the answer accuracy and reliability essential to legal practice" — specifically, that hallucinations are suppressed by generating answers only from trustworthy sources, and that citations are explicit enough to make verification easy.

Going forward, the company is running a parallel proof of concept for multi-agent coordination, and plans to move to a seamless usage environment via Mirai Translate's AI platform.

How it differs from other legal tech

Legal tech divides by what it makes more efficient. Legal Brain is centered on research — the process of finding out — which is a different scope from products built primarily around contract review.

In the same research space, Legalscape offers a service built around a legal literature database, where the range of documents covered and the search experience are the main points of comparison. For contract review, LegalForce, OLGA, and LeCHECK apply; because their purpose does not overlap with the "finding out" that Legal Brain handles, they are often used together.

Information on this page is based on what was published on the official site and in press releases from Bengo4.com, Inc. as of August 2026. Please check the official site for the latest specifications and pricing.