Source Document Summary

Overview

Fab Forward (株式会社ファブフォワード) established an anti-social forces screening program (反社チェック) in November 2025. The program ensures all new business partners — corporations, sole proprietors, their representatives, and officers — are screened before transactions begin. The following is a summary of the four internal policy documents and the technical research paper that form the basis for this application.

Document ①: 反社会的勢力チェック基準 (Screening Criteria)

Purpose: Exclude anti-social forces from all business relationships.

Definition of anti-social forces: Yakuza (暴力団), associates, front companies, corporate extortionists (総会屋), political racketeers, and any closely affiliated parties.

Check targets:

• All new counterparties (corporations, sole proprietors)
• Representatives and officers
• Where necessary: major shareholders, beneficial owners, subsidiaries, parent companies, key trading partners

Check methods (to be used in combination):

1. Anti-social force databases (police publications, commercial investigation firms)
2. Public records — official gazette (官報), registries, newspaper articles, court records
3. Written pledges / questionnaires from counterparties
4. Internal records (past incident logs)
5. Credit bureaus (TDB, TSR, etc.)

Rejection criteria: Confirmed anti-social affiliation, known relationships, influence by anti-social forces, refusal to submit pledges, false declarations.

Ongoing obligations: Annual re-checks, ad-hoc checks on major incidents, 5-year data retention after relationship ends.

Organizational roles: Sales (request), Legal/General Affairs (investigation), Compliance Officer or CEO (final approval).

Document ②: 反社チェック事務フロー (Operational Flow)

Three-track flow based on company type:

Nikkei Telecom search: Search company name, representative, and each officer against a fixed set of ~50 negative keywords (organized crime, arrest, fraud, drugs, etc.) across Nikkei newspapers, national papers, and general press.

Internet search: Same keyword pattern applied via web search engine for each entity.

Critical cost rules:

• Group subsidiaries perform searches only — they do NOT click/view articles (clicking incurs charges)
• If hits are found, subsidiaries notify FF Legal, who performs the deeper article review
• TDB data retrieval costs (¥1,600/case) are charged back to each subsidiary monthly

Document ③: 反社チェック事務取扱 (Operational Procedures)

Step-by-step workflow:

1. Sales initiates: Fills "取引先チェックシート" (counterparty check sheet), submits to Legal
2. Legal records: Enters search date, company name, representative, address into "反社チェック検索結果一覧表"
3. Company identification:
   • Listed → verify via EDINET
   • Unlisted with officer info → proceed to Nikkei Telecom
   • Unlisted without officer info → retrieve from TDB COSMOS2 (cost-incurring, last resort)
4. Nikkei Telecom search (detailed steps):
   • Log in, navigate to 記事検索 (article search)
   • Pre-saved search condition "反社チェック" loads the ~50 negative keywords
   • Execute base search, then use "絞込み" (narrow down) for each entity name
   • Each entity must start from a fresh search — narrowing from a previous entity produces incorrect results
5. Internet search: Same keyword set via web search engine
6. Result handling:
   • 0 hits on both → mark "no concerns", record checker name, return check sheet to Sales
   • 1+ hits on Nikkei Telecom → notify FF Legal (do NOT view articles), forward check sheet
7. FF Legal review: Logs into Nikkei Telecom, reads flagged articles, records findings on check sheet, returns to subsidiary
8. Recording: All results recorded in 検索結果一覧表; rejection reasons documented in remarks

Document ④: 反社チェック検索結果一覧表 (Search Results List)

A tracking spreadsheet with columns:

• No., Search Date, Target Company Name, Representative Name, Target Company Address
• Search Results, Investigation Confirmation, Comments (rejection rationale)
• Rows for company name, representative, and up to ~50 officers per entity

Both Nikkei Telecom AND internet search must be performed for every entity.

Research Paper: Technical Implementation Analysis

Key thesis: A 3-tier screening architecture optimizes cost, speed, and accuracy.

Tier 1 — Zero-cost automated OSINT screening (internal module)

• Government open data: 官報 (official gazette), 国土交通省 negative info system, 金融庁 administrative actions/license lists
• 法人番号 (corporate number) API from National Tax Agency as the primary key for entity resolution
• Real-time, millisecond-level checks at zero marginal cost
• Auto-block on exact matches (e.g., unlicensed operators, sanctioned entities)

Tier 2 — Paid API high-precision filtering (external vendors)

• Commercial services: RISK EYES (¥300/query), RiskAnalyze (¥116-275/query at scale), RoboRobo, G-Search, Alarm Box
• Decades of curated media articles, proprietary noise-filtering algorithms
• Reduces human review to the truly suspicious few percent

Tier 3 — Deep investigation via specialized sources (existing resources)

• TDB credit reports and Nikkei Telecom for final-stage high-value decisions
• Requires compliance officer approval workflow
• Manual, high-cost, reserved for gray-zone or large transactions

Cross-cutting concerns

• Entity resolution: Corporate number (法人番号) as hub — company names change frequently, especially for front companies
• Legal compliance: Scraping permissible under Copyright Act Article 30-4 (information analysis exception) for fact extraction, but must respect robots.txt and site ToS
• Audit trail: All check results must be persistable as PDF/text for future IPO audits or regulatory inspections
• False positives: Google search automation produces noise (defamation, competitor sabotage, homonyms) — must be treated as preliminary alert only, not definitive judgment