Inside Japan's ¥15 Billion Disability Welfare Fraud: Why Type-A Employment Support's Design Enabled the Abuse
In March 2026, Osaka City revoked the licenses of four Type-A continuous employment support offices operated by Kizuna Holdings and demanded over ¥11 billion in refunds. The total nationwide fraud was certified at approximately ¥15 billion. A scheme internally known as the "36-Month Project" cycled recipients of the "Employment Transition Support Structure Addition" to multiply additions. Roughly 100x the scale of the 2017 Ajisai no Wa case, this exposes the structural flaw of a reward system where monetized outcome metrics make falsification economically rational.
TL;DR
- On March 27, 2026, Osaka City revoked licenses of four Kizuna HD offices. Refund demand totals ¥11.08 billion; nationwide fraud certified at ~¥15 billion.
- The scheme cycled recipients between Type-A status and group-company "general employment" every few months to repeatedly claim the Employment Transition Support addition.
- At ~100x the scale, it repeated the structural pattern of the 2017 Ajisai no Wa case (~¥137 million, 280+ dismissals).
- The 2024 fee-schedule revision tightened criteria but triggered 329 Type-A closures and ~5,000 dismissals between March and July 2024.
What Is Happening
The facts of Osaka City's license revocation of four Kizuna HD offices and the ¥11 billion refund demand
On March 27, 2026, Osaka City issued an administrative license revocation under the Comprehensive Support for Persons with Disabilities Act against four Type-A continuous employment support offices operated by Kizuna Holdings. The targeted offices — Lian Uchihonmachi, Liberara, Reve, and Mirrime — will close at the end of April 2026, with the revocation taking effect on May 1.
The certified fraudulent claims amount to approximately ¥7.9 billion for Osaka City alone, with a refund demand of ¥11.08 billion after a 40% surcharge. Including approximately ¥7.1 billion of fraudulent claims submitted to 75 municipalities across 7 prefectures, the nationwide certified fraud totals approximately ¥15 billion, involving more than 14 prefectures and 104 municipalities.
This article distinguishes clearly between the two figures. ¥11 billion is Osaka City's refund demand (¥7.9 billion principal plus the 40% surcharge); ¥15 billion is the national total certified by all affected jurisdictions. For a single operator group, this is an unprecedented scale in the history of Japan's disability welfare service system.
Background and Context
The structure of Type-A continuous employment support and the fee-schedule history since 2017
What Type-A Offices Are
Type-A continuous employment support is a disability welfare service under Article 5, Paragraph 14 of the Comprehensive Support for Persons with Disabilities Act. Offices offer employment opportunities and training under formal labor contracts to persons with disabilities who face difficulty in ordinary employment. The decisive difference from Type-B is the presence of the labor contract: Type-A users are legally "workers" under the Labor Standards Act and are covered by minimum wage law.
Operators must pay at least the prefectural minimum wage, and the funding source must in principle come from productive-activity income. This is the unique operational pressure of Type-A. Offices that cannot cover wages from productive-activity income will, in effect, divert training-etc. benefits (the service fee) to cover wages — a deviation from the program's intent. The Ministry of Health, Labour and Welfare codified this principle in its March 30, 2017 notification, but as of March 2023, 50.7% of offices failed to meet the designation criteria.
The Two-Tier Reward Structure
The Type-A fee consists of a base fee plus additions. The base fee was shifted in the 2018 revision to a seven-tier system keyed to average daily working hours, then restructured in 2021 into a five-item scoring system (working hours, productive activity, diversity of work styles, support capacity, regional collaboration). The 2024 revision expanded this to seven items and introduced a 20-point penalty for offices whose productive-activity balance fell below total wages for three consecutive periods.
Among the additions, the Employment Transition Support Structure Addition lay at the core of this case. This addition adds points to total user days for the current fiscal year based on the number of people who moved from Type-A to general employment in the previous year and maintained that placement for at least six months. According to snabi-biz, the design intends to incentivize operators to produce genuine general-employment transitions. The logic is coherent — but the counting unit is "persons × days," and nothing in the rule forbade operators from hosting those "general employment" positions inside their own corporate group. That was the gap.
The Ajisai no Wa Precedent (2017)
In July 2017, a general incorporated association called "Ajisai no Wa" in Okayama and Kurashiki closed its offices, dismissing more than 280 persons with disabilities at once. In September, executives were referred to prosecutors on suspicion of fraud, and damages from inflated working-hour records were certified at approximately ¥137 million. It was also reported that subsidies had been diverted to eel farming and yakiniku operations.
In response, the Ministry of Health, Labour and Welfare revised the designation criteria in March 2017, codifying the principle that wages must come from productive-activity income, and introduced the business-improvement plan system. Nine years later, the Kizuna HD case replayed the same structure at roughly 100x the scale. The difference is that Ajisai no Wa was a "failed fraud" — exposed by bankruptcy — whereas Kizuna HD was a "successful fraud," undetected for more than two years.
Reading the Structure
How addition cycling became economically rational, compared with the 2017 Ajisai no Wa case
The Mechanics of the "36-Month Project"
The scheme, reportedly known internally within Kizuna HD as the "36-Month Project," cycled the Employment Transition Support addition. A note of caution: "36-Month Project" is not a policy term but Kizuna HD's internal nomenclature.
The cycle works as follows.
- A user of a Type-A office (already under a labor contract) is re-employed at a group-affiliated company as "general employment."
- After six or more months, they are counted as a "retained transition," and the addition begins to be claimed from the following fiscal year.
- The same person is later returned to Type-A user status.
- They are then moved to another group-affiliated company for another round of "general employment," again becoming eligible for the addition after six months of retention.
- Repeating this on a roughly three-year (36-month) cycle stacks multiple claims per person.
Osaka City's certification language is explicit: "the office systematically manufactured an excess of retained transitions by repeatedly moving users into self-employed positions under the operator's direction" and "the internal employment is merely a step in the support-plan process and does not qualify as a continuing support structure aimed at employment retention." The Ministry had stated in the 2024 fee-schedule revision that "claiming multiple times for the same user within a three-year window is not envisioned," yet the Kizuna group reportedly continued its claims after the revision.
The Hollowing of Licensing Review and Cross-Jurisdictional Detection
In its September 2024 policy request, the National Council of Social Welfare Employment Centers criticized the fact that certain licensing authorities (prefectures, designated cities, and core cities) were operating on a "grant the license if the papers are in order" basis. The fact that over half of all offices fell short of the criteria more than seven years after the 2017 reforms demonstrates the gap between document review and on-site inspection. In the Kizuna HD case, claims reportedly continued even after repeated city guidance, and the slow escalation from guidance to audit to sanction is what allowed the damage to balloon.
A second structural challenge is cross-jurisdictional detection. National Health Insurance Federation claim data are aggregated prefecture by prefecture and municipality by municipality, and there is no robust integrated mechanism to detect fraud patterns across multiple jurisdictions. That Kizuna HD's fraud spread across more than 14 prefectures and 104 municipalities suggests that per-municipality claim amounts may have been designed to stay below local audit thresholds.
The 2024 Revision Domino and the Metrics Paradox
The 2024 revision strengthened penalties on offices unable to fund wages from productive activity. The consequence: between March and July 2024, 329 Type-A offices closed nationwide; Hello Work recorded 4,884 dismissals, of which 4,279 were Type-A users. As of the end of August 2024, re-employed individuals numbered 936 and planned transitions to Type-B totaled 2,073, leaving many users without a clear destination.
Here lies the essential dilemma of program design. Tightening rules can rationally remove operators who cannot function as employers, but it simultaneously strips users of their destinations. And it cannot, by itself, exclude operators like Kizuna HD who turn tightening into a more lucrative fraud target. Outcome metrics in addition-based systems carry an inherent paradox: the more they are monetized, the more economically rational their falsification becomes.
The Commodification of Welfare Labor
Within Japan's post-ratification (2014) disability policy framework centered on "deinstitutionalization, community transition, and general employment," Type-A was designed as an intermediate good between institutional welfare and ordinary employment. The number of Type-A offices grew by roughly 15% from 3,842 in 2019 to 4,415 in 2024, with users rising about 18% from 72,197 to 85,421. For-profit entry drove this growth, and within it the Kizuna HD case shows what happens when the "general employment" label is consumed as a generator of additions.
Within the cycle, the user's decision-making, career formation, and retention support were hollowed out; users were treated as "accounting units that generate addition fees." When program design is flawed, the ideals of deinstitutionalization can be transformed into a "revenue device in a different shape." This case marks one critical threshold in the commodification of welfare labor, and compels us to ask who employment support is really for. This article keeps its subject on "offices," "operators," and "program design" precisely to avoid treating users as mere victims within the program's objects.
Further Reading
For readers seeking a systematic understanding of Type-A Supported Employment program design and disability employment practice, Kaiteiban Shōgaisha Koyō no Jitumu to Shūrō Shien (Revised Edition: Practice and Employment Support for Persons with Disabilities) (Tomoko Mashino, Nihon Hōrei) — available in Japanese — covers the legal framework, benefit structures, and support operations in comprehensive detail.
References
Osaka City Revokes Licenses at 4 Disability Employment Offices, Demanding ¥11 Billion Refund (2026)
Notice of March 27, 2026 (Regarding the License Revocation) (2026)
Notification on Appropriate Operation Standards for Type-A Continuous Employment Support (2017) (2017)
Policy Request Regarding Type-A Continuous Employment Support Offices (September 2024) (2024)
Ministry Clarifies Response to Type-A Closure Cases in Notice to Local Governments (2024)
The ¥15 Billion Fraud Case at Kizuna Holdings Group (2026)
Mass Dismissal of Persons with Disabilities — Families Voice Anxiety and Outrage (Okayama) (2017)
Understanding the FY2024 Fee Schedule Revision for Type-A Continuous Employment Support (2024)