Most AI answers about a Japanese company are summaries. Someone, or something, has already decided which parts of the annual report matter and in what words. For the section of the report where the company itself lists what could go wrong, that is the part you want to read in the company's own words.

Kabuken now has a tool for that: get_filing_text. It returns two narrative sections of an annual securities report (有価証券報告書) as the company filed them on EDINET, the Financial Services Agency's disclosure system: 事業等のリスク (business risks) and the management's analysis of financial position, operating results and cash flows (MD&A). The text is Japanese, taken verbatim from the filing with markup removed. No AI processes it, and the response names the XBRL element it came from.

One worked example

Nireco Corporation (株式会社ニレコ, EDINET code E02411, securities code 6863) makes control, measurement and inspection equipment. We asked for the business risks section of its annual report for the fiscal year ended 31 March 2026. The tool call is:

get_filing_text
  entity_code: E02411
  sections: ["business_risks"]

The response identifies the filing as EDINET document S100YFWS and names the source element:

"section": "business_risks"
"source_element": "jpcrp_cor:BusinessRisksTextBlock"
"label_ja": "事業等のリスク"
"total_chars": 2488
"truncated": false

The text itself starts as the filer wrote it:

３【事業等のリスク】

有価証券報告書に記載した事業の状況、経理の状況等に関する事項のうち、投資者の判断に重要な影響を及ぼす可能性のある事項には、以下のようなものがあります。

The section then runs through nine numbered items, from ① 経済状況による業績への影響について and ② 競合に関するリスク to ⑧ 自然災害に関するリスク and ⑨ 退職給付債務. All of it comes back in one page because the section is shorter than the default page size.

Two things are worth noticing. First, the headings and their order are the company's, not a model's. Second, source_element tells you which block of the XBRL filing the words came from, so you can find the same passage in the original document.

To try it yourself, ask:

Show me the business risks (事業等のリスク) section of Sony Group's latest annual securities report, in the original Japanese.

How it works

• Give either entity_code (the tool uses the company's newest annual report that has XBRL data) or doc_id (a specific filing).
• sections is business_risks, md_and_a, or both. The default is business_risks.
• Long text is paged by characters. max_chars sets the page size (1 to 30,000, default 12,000); send the returned next_cursor with the same arguments to continue.
• If a filing does not have a section stored, the tool lists it in missing_sections instead of guessing.

Step-by-step guide: https://kabuken.intellics.ai/docs/en/guide/filing-text/?utm_source=blog&utm_medium=article&utm_campaign=filing_text

Full input reference: https://kabuken.intellics.ai/docs/en/api/tools/?utm_source=blog&utm_medium=article&utm_campaign=filing_text

What the tool does not do

The tool does not summarise, translate or interpret. The text comes back in Japanese. In a chat, your AI assistant reads it and can translate or summarise it for you, but that translation or summary is the assistant's work, not the filing's. If a sentence matters, check it against the quoted original and the document ID.

Kabuken has no stock prices and gives no investment advice. The risk section is the company's own statement of what could affect its business; this article does not rate any company or say whether the risks matter.

Which accounts can use the tool, and the daily lookup limits, are set out in the tool reference linked above.

Try it

Connect Kabuken to Claude or ChatGPT, then ask in your own words, for example: "Show me the business risks section of Nireco's latest annual report, exactly as filed." Setup guides: https://kabuken.intellics.ai/docs/en/?utm_source=blog&utm_medium=article&utm_campaign=filing_text

In the plugin: version 2.1.9 of the Kabuken plugin adds a filing-text skill. It tells the assistant to use get_filing_text when you ask for risk factors, 事業等のリスク, management discussion, or the original text as filed. Version 2.1.9 is available in the ChatGPT plugin directory and from the GitHub repository: https://github.com/Intellics-ai/kabuken-mcp-plugin . In Claude, the Kabuken MCP server works through the connector; the tool is the same.

Kabuken: https://kabuken.intellics.ai/?utm_source=blog&utm_medium=article&utm_campaign=filing_text

Source: EDINET (Financial Services Agency, Japan), Public Data License 1.0. Processed by Kabuken. The text of a filing belongs to the filing company; it is reproduced verbatim and was not edited by Kabuken. Takedown requests: https://kabuken.intellics.ai/support?utm_source=blog&utm_medium=article&utm_campaign=filing_text