EOS is dead, long live ITE: the operational arm of Dutch tax profiling

Summary

Anyone reading the documents about the Belastingdienst scandals keeps stumbling over three letters: EOS. The External Government Cooperation team ran the mailbox where tips and informant letters arrived, decided which signals were investigated and, according to the Fraud Enforcement Plan, drafted so-called pre-weighing documents for the RIEC, the regional partnership against organised crime that undermines society. In the official papers about the period after, EOS quietly disappears. The name is not a typo and the team was not abolished. It was absorbed into something bigger.

The bigger thing is called ITE: Intensive Supervision and External Government Cooperation. Eleven teams across the country with their own auditors, assessment officers, collectors, penalty coordinators and a liaison officer for criminal cooperation. In 2026 more than half a million hours are reserved for these teams, including 131,400 hours of tax audits. They run cases for the RIEC, the Public Prosecution Service and the FIOD. The FIOD posts VAT fraud signals directly to them. And the department that once built the banned risk model RAM, DF&A, is developing new models for the ITE work process right now.

This article reconstructs the transition from EOS to ITE from public records: the work and production plans of the Belastingdienst, a Woo decision about tips and informants, the sweep operation report on unlawful lists and parliamentary documents about RAM and its successors. The outcome fits in one sentence. Where three control systems went under in 2020, the Belastingdienst built a single organisation that unites selection and enforcement.

Three systems fell in 2020, one organisation came back

To understand the transition you need the context of 2020. Three Belastingdienst control techniques reached their end that year.

The first was RAM, the Risk Analysis Model. For twenty years this system had linked records of an estimated eleven million citizens and 1.9 million businesses in a data warehouse of 250 tables, fed by 69 source systems. No legal basis, no notification, no right of objection. The KPMG investigation of February 2025 and the parliamentary letter of 6 March 2025 confirmed what critics had suspected for years: the system was structurally at odds with statutory requirements. RAM went dark on 25 May 2018, one day before the GDPR took effect.

The second was the FSV, the Fraud Signalling Facility, the blacklist system in which an estimated 180,000 citizens were registered as fraud signals. The Dutch Data Protection Authority fined the Belastingdienst 2.75 million euros over it in 2020.

The third fell more quietly. On 2 July 2020 the processing of tips and informant letters was halted. This is stated verbatim in the SME work and production plan 2026: “The process for handling signals, i.e. tips/informants, has been suspended since 2 July 2020.” A year earlier, Detection and Selection at the Gate had been ended, the scheme under which entrepreneurs were selected at the core. A compensation scheme now exists for the latter: the Compensation for Selection at the Gate Act of 19 December 2023.

Three shut-down systems. None of them really disappeared. RAM gained successors that the Data Protection Authority ordered to be phased out immediately in July 2025, in vain. FSV data seeped into successor systems. And the tips process was not abolished but rebuilt, with a new organisation around it.

What EOS was

The oldest public trace of EOS lies in a Woo decision the Belastingdienst issued on 6 March 2023 in response to a request from September 2021. A requester had asked for all documents on the handling of tips and informant letters in the broadest sense, from 2005 through 2021. The decision with annexes, published via open.overheid.nl, shows how that process was set up.

A tip could arrive by phone call to the tax line, an email to the FIOD or a handwritten note pushed through the letterbox of a tax office. Everything landed in a dedicated mailbox. That mailbox was managed by SME staff of team EOS, External Government Cooperation. They made the first cut: which signals the SME directorate would pick up itself and which would go to the individuals department. The Fraud Enforcement Plan among the annexes budgeted roughly a thousand signals for 2017, under project code 1194.

The Woo decision also contains a passage from a Fraud Enforcement Plan that shows EOS to the outside world. Under the stakeholders heading, the division of labour reads:

SME, Team EOS and RIEC: supplying tips and informant signals, drafting and supplying pre-weighing documents to the national liaison officer.

That single line carries more weight than it seems. The RIEC, the Regional Information and Expertise Centre, is the partnership in which police, prosecutors, municipalities, the FIOD and other services jointly “weigh” organised crime cases: they decide together which file goes to which track, administrative or criminal. EOS was already delivering the Belastingdienst contribution to that weighing, in the form of ready-made pre-weighing documents. The link between fiscal selection and the undermining apparatus is therefore not recent. It was in the enforcement plan back in 2017.

The presentation of 28 November 2017: the face of EOS

How EOS worked on the inside is documented in a presentation the team itself gave: “Toeslagen — external partners”, dated 28 November 2017, drafted by the Enforcement Steering Intelligence Department with the External Government Cooperation Team as its author (document number 00927, reference 1874992). The document is published online by the Childcare Allowance Information Point (pdf, 27 MB) and confirms in one stroke what this reconstruction assembled from loose pieces.

It confirms, first, that EOS was no SME sideshow but the presenter of the Benefits enforcement machine itself, to external partners. The slides name the RIEC and LIEC, the Combined Team Approach to Facilitators “with mandated Belastingdienst colleagues”, the LSI, the National Approach to Address Quality, the VNG, KCH and the FIOD, with instruments such as home visits and covenants, deployment of GOB and ANPR, area-based approaches to risk addresses and holiday parks. One slide notes that as of 1 January 2014 the Benefits department joined as well — in the middle of the period later known as the allowances scandal.

The presentation then shows the profiling in full colour. The enforcement vision reads “Everyone gets supervision”, elaborated with risk classification “as early as possible in the chain”, IP monitoring during the application and provisional award, and dashboards plotting the whole country: each dot a postcode area, its size the number of applicants, its colour the number of assigned risk points per thousand applicants. Risk scores by postcode, in 2017, while the since-banned RAM had been switched off for less than a year and the sweep operation would later find 35 RAM spreadsheets with postcode selections.

And the later ITE architecture is already there. The organisation slide: “3 teams for intensive subject-oriented supervision, 2 teams for processing fraud signals, penalty-fraud coordinators, liaison officer” and a prosecutor specially assigned to Benefits. Placed next to the ITE vacancy text — audit, assessment, collection, information desk, penalty-fraud coordinators, liaison officer — it is the same kit of parts, four years before ITE was founded. The team’s contact line ran through its own mailbox: toeslagen-externeoverheidssamenwerking@belastingdienst.nl.

The presentation of 28 November 2017 was not a one-off either. The same release by the Childcare Allowance Information Point contains a whole series: document 00926 is a “RIEC presentation” of 15 November 2015, two years before the external-partners version; the archive holds follow-up presentations for the RIEC of 28 January 2016 and 13 April 2016; and eight days after the 28 November 2017 presentation, virtually the same package stood before the Haaglanden regional police, as the “Toeslagen Gouda presentation” of 20 November 2017. The pattern is a circuit: from RIEC tables in 2015 and 2016 via external partners to the police in 2017, each time with risk selection, intensive supervision and foreign-nationality signalling on the slides.

The sweep operation that justice and the Belastingdienst conducted after the allowances scandal in search of unlawful lists also found EOS traces. The Prio 1b report of 18 February 2022 and its appendix inventoried 194 lists containing nationality, medical or criminal records. Among them: two lists named “LOA Supervision fraud eos” (risk rating low), a list “FIOD EOS Audit” (medium risk), a list “Fraud EOS”, explained with “EOS stands for external government cooperation” (fell outside the population), and an “eos forestry project based on risk selection” (no risk). Four of the six EOS lists simply survived the risk assessment.

What ITE is

The current form is called ITE, Intensive Supervision and External Government Cooperation. The definition sits in the organisation’s own vacancies. In 2026 the Belastingdienst recruited supervisory officers for the teams in Arnhem, Rotterdam, Hoofddorp and Maastricht, describing the work like this:

Within the ITE teams, cooperation takes place in the field of intensive fiscal supervision and external government cooperation. These teams steer the approach to suspected fiscal misconduct and work with other organisations on tackling undermining crime. The latter happens among others in RIEC partnership with municipalities, the Public Prosecution Service, police, the Royal Marechaussee, the FIOD and Customs.

The vacancy counts eleven ITE locations. The SME work and production plan 2026 speaks of twelve reporting desks within the ITE teams. A team is not a letterbox but a complete enforcement outfit: besides the manager it holds audit staff, income and payroll assessment, VAT assessment, collection, the information desk, penalty-fraud coordinators and a liaison officer under the General Act on State Taxes, the link to criminal law. The vacancy names the cases explicitly: entrepreneurs “suspected of fiscal misconduct”, often in the context of organised undermining crime, across every tax type.

Inside the Belastingdienst, ITE cases occupy a special position. The 2026 plan describes the allocation framework that distributes tax returns over available capacity. At the top, as the first category before treatment intentions and automatic corrections, sit the “risky networks and returns selected for intensive supervision and external government cooperation (ITE)”. Whoever is flagged by ITE stands at the front of the queue. That is no longer ordinary auditing but a coordinated campaign.

The transition, year by year

The reconstruction from the documents shows a calm, almost administrative transition. That is precisely what makes it striking.

Read the table as one movement. The old name vanishes from the documents between 2023 and 2025. In the plan for the second half of 2025 the category is still called “Intensive Supervision and EOS”, as if the organisation itself was getting used to the new name. In the 2026 plan EOS is permanently replaced by ITE. Nowhere is there talk of cutbacks. The number of places where this work happens grows from one mailbox to twelve reporting desks.

The session minutes of July 2022, released through the Childcare Allowance Information Point, show that the culture moved along with it. One participant argues that an auditor who comes across an interesting fraud file should not hand it over but “become part of the investigation group within ITE or get involved in the criminal investigation”. The hinge between fiscal supervision and criminal prosecution is discussed there as ordinary working practice.

The RIEC machine

The centre of gravity of ITE lies with the cooperation that once ran through EOS and the pre-weighing documents. The 2026 work plan lists the partnerships the teams sit in, and it is a long list.

The RIECs themselves, and their national counterpart LIEC, for “jointly giving substance to an administrative and integrated approach to organised undermining crime”.

The birthplace: from TaskForce B5 to two RIECs

Before the RIECs existed, there was Brabant. In 2011 the minister of Security and Justice, at the request of the local administration in Brabant, established TaskForce B5, continued in early 2014 as TaskForce Brabant Zeeland (BZ). The aim: “disrupt the criminal industry more effectively”, focused on cannabis, synthetic drugs and outlaw motorcycle gangs, and above all on “increasing the operational integrated clout”. The 2014 interim tally: 23 criminal networks tackled (2012: 11), seizures by the Integral Seizure Team worth 8.3 million euros (2012: 6 million), 99 building closures and 340 Bibob applications across twenty municipalities. Justice contributed half a million euros in 2014, plus 750,000 euros for the seizure team. This taskforce model — an integrated information position, fiscal and administrative interventions alongside criminal law — became the blueprint for the national approach. After an intermediate phase as Taskforce-RIEC Brabant-Zeeland (2019 to early 2025), the partnership has existed since 2025 as two full RIECs: RIEC Oost-Brabant and RIEC Zeeland-West-Brabant.

The same parliamentary debate of November 2014 also contains the seed of the current legal foundation. The minister announced “an exploration of a framework act for data exchange” together with his fellow ministers, and a national roadshow in which the leadership of justice, police, the LIEC and the Ministry of Finance with the Belastingdienst toured all regions. That 2014 exploration led to the Data Processing by Partnerships Act of 1 March 2025. The line of this article thus runs from 2011 to 2026 without a break: taskforce, RIEC, WGS, ITE — each time a new coat around the same integrated approach, with the Belastingdienst at the table from the start.

What that looked like operationally is documented in a taskforce progress report covering June to November 2015. The taskforce presented its “nine-pack” — vital elements of the criminal industry, from kingpins and assets to facilitators, cutting rooms and transporters — and its approach to “untouchables”: “show-offs or types deliberately operating under the radar” tackled integrally. About that group the report notes, without any hedging:

Under the flag of the RIEC, police, the Public Prosecution Service, the Belastingdienst and municipalities are currently monitoring these subjects. Wherever they slip up, they are tackled. This partly on the basis of an integral intervention matrix.

Subjects were thus already being monitored by the Belastingdienst in RIEC context in 2015, before any criminal offence, with an intervention matrix at the ready. The same report describes targeted fiscal audits of businesses behind growshops, an integral seizure team that was “carved up” in 2014 and disappeared into the undermining teams, only to return in Zeeland-West-Brabant “reinforced and separately” with “an impulse” for the Belastingdienst’s involvement, and the South Netherlands Intensification that started on 18 November 2014 with 125 specially labelled police officers and 15 prosecutors, later joined by the FIOD and the Belgian authorities. A footnote also reveals why fiscal seizure results are barely verifiable: the Belastingdienst’s systems “do not provide information on specific subpopulations”. Anyone wanting to know what the fiscal column of this approach yielded cannot check it back — a data gap that still exists in 2026.

Within that system, alongside prosecuting individual offenders and dismantling networks, administrative and fiscal enforcement are central. The partnerships in which the ITE teams are anchored range widely. Then the National Cooperation Alliance for Compliance with the Social System and Labour Laws, LSN, formerly LSI: Belastingdienst, UWV, SVB and the Labour Inspectorate together against abuse of benefits, premiums and labour laws, under Article 64 of the SUWI. Then iCOV, the Infobox Criminal and Unexplained Assets, to which the Belastingdienst supplies two products under a delivery decree: the iCOV Relations Report and the iCOV Assets and Income Report. Then the Strategic Knowledge Centre, where police, prosecutors, FIOD, Customs, Marechaussee, municipalities and private parties produce threat assessments, with a liaison that ITE itself seconds. Then the approaches to misconduct in the cleaning and temping industries, AMS and AMU, to which an international component was added on 1 January 2024 following a recommendation of the Roemer committee. Then the Risky Networks team in Zwolle, which draws up treatment plans per network. And then the Care Fraud Node, where the SME directorate sits in through ITE’s national coordination point.

The production plans give the scale. For 2026, ITE stands for 562,200 hours, of which 131,400 hours of tax audits, 47,100 hours of fraud enforcement and 37,800 hours for the information desks. The 2025-2 plan books a production planning of 67,500 hours for the RIEC clusters. The SME directorate’s regular audits are not included: for those, the 2026 target is 7,000 to 10,000 audits, excluding ITE. ITE audits come on top.

Some cases reach the teams by a remarkably direct route. For VAT carousel fraud the Belastingdienst runs a detection tool that sorts signals into four priorities. Priority 1 signals, the missing traders, and priority 2, intra-community fraud, are handled within the ITE teams. And the VAT Fraud Coordination Point, part of the FIOD, “posts signals and orders directly to the SME ITE teams for handling”. The criminal investigation service orders, enforcement executes. For used car imports, ITE staff likewise conducted the first BPM audits in 2025, with several cases expected to be reported “under the AAFD protocol” for criminal investigation.

The legal foundation followed afterwards, but what the act actually did deserves a closer reading. On 1 March 2025 the Data Processing by Partnerships Act entered into force, with an accompanying decree. The act designates four partnerships: the Financial Expertise Centre, iCOV, the RIECs (the work plan also names the LIEC) and the Care and Safety Houses. The implementation track “mainly touches the work within the ITE teams”, and an implementation test by the Ministry of Justice and Security follows in March 2026.

Did this act make illegal profiling legal? Not for the past: RAM and the FSV remained unlawful and the act has no retroactive effect. But it did legalise the channel. Article 1.5 obliges every participant to supply the categories of data designated by decree to the partnership — an obligation, not a power — and for that purpose overrides confidentiality provisions insofar as they admit an exception. Article 1.6 turns the purpose of the partnership, an “objective of substantial public interest”, into the processing basis itself. And the definitions in article 1.1 build profiling in: “steering information” expressly includes “the underlying data from a risk analysis”, and an “enforcement bottleneck” can be a person, a group, an area or an entire sector. The act claims to regulate data processing; substantively it regulates risk-driven joint file building.

Are all personal data now shared? Within the four partnerships: virtually. The designated categories cover personal details with citizen service numbers, asset components, relationships and contacts, “earlier unlawfulness” and police and judicial data; article 1.5, third paragraph, explicitly covers criminal record data, and article 2.21 even allows processing of special categories (article 9 GDPR) and the citizen service number for case consultations in RIEC context. The safeguards are real but narrow: human intervention before results of automated analysis are shared, public explanation of the patterns used, deletion within five years, and no tax data to private participants. Critics saw more shadow than light: the Netherlands Jurists’ Journal noted that the partnerships “operated without a clear legal basis before the act entered into force” — the act thus legalised existing practice — and the Council of State, the Data Protection Authority and commentators flagged tension with privacy, fiscal confidentiality and effective legal protection. The core caveat for this investigation: the safeguards of article 1.9 apply only within the WGS partnerships. The VVG and FAM that DF&A is building for the ITE teams fall outside them. Internal Belastingdienst tools face no human-review test and no public explanation duty.

The DPIA for the RIEC cooperation, for which a process was already set up in 2021, is being revised in 2026. And about the old LSI covenant the work plan notes that the Data Protection Authority “some years ago” flagged risks in the description of joint controller responsibility, prompting process descriptions, a privacy protocol, an amended covenant and new DPIAs. What the legislator has given a basis for since 1 March 2025 had been running for years.

The teams and the officeholders

The Belastingdienst is reticent about who exactly works in the ITE teams, but vacancies, the 2026 Annual Plan and organisational publications sketch a clear line, from the field to the cabinet.

The vacancy counts eleven locations; the 2026 work plan counts twelve reporting desks. Nationally the whole is coordinated by the national coordination point (LC ITE), which also holds the Belastingdienst seat in the care fraud cooperation group and manages the BOA line.

The vacancy for a regional director in Hoofddorp describes how far the entanglement reaches. The regional director “represents SME in the RIEC, a partnership of various authorities such as Customs, FIOD, Public Prosecution Service and police that combat undermining crime together”, and must bring “affinity with intensive supervision and external government cooperation (ITE)”. The administrative and the operational RIEC line thus converge in a single post. The same picture holds on the partner side: RIECs are chaired by mayors, and municipal contact sheets list “team leader ITE Belastingdienst” and “account manager/head of RIEC” as permanently paired contacts.

The timing of the appointments tells its own story. Marjan Willemsen started as director in December 2021, the year ITE got its information desks. Berry Roks left for the Individuals directorate on 10 September 2026, the year the work plan made ITE the only name. His old seat has been empty since; the recruitment text explicitly asks for affinity with intensive supervision and external government cooperation.

The thread to RAM

Anyone searching the post-2021 documents for profiling will not find it in the name but in the function. The connection between the old system and the new runs along three threads.

The first thread is the outlet. RAM had a traffic light mechanism: red meant selection for the supervision or fraud track. Our earlier RAM investigation describes how the top 20,000 red subjects automatically proceeded to the tax intermediary approach, the fraud/EOS track or thematic investigations, without substantive review. EOS was no sideshow in that system but the destination. The entrepreneur pointed out by the traffic light was, in practice, approached by the line now called ITE.

The second thread is the builder. RAM and the SME selection layer RDM were built and maintained by department DF&A, as KPMG noted in section 6.6.10 of the RAM report: RDM drew its information from RAM. Logging from 2017 showed that 89 percent of RAM usage sat with the SME directorate, good for 19,091 queries in a single year. The same DF&A, according to the 2025 and 2026 work plans, is “in cooperation between the ITE teams and DF&A” building the new models for the ITE work process. The stack of index cards was replaced. The builder stayed.

The third thread is the recipient. RAM and RDM data was shared through covenants with ten external partnerships, including the RIECs, the LIEC, the LSI, iCOV and the CT infobox. The parliamentary letter of 6 March 2025 identified, among 1,170 RAM extracts found, 14 spreadsheets with selection on nationality and 35 with selection on postcode. The channel through which profile data reached the undermining apparatus runs through exactly those partnerships where the ITE teams form the Belastingdienst side. And the channel is still open: the Information Template procedure, with which staff hand over personal data in Excel format to police, prosecutors, RIEC and iCOV, covered roughly 38,000 subjects in 2024. The Data Protection Authority demanded immediate replacement in July 2025. The Belastingdienst’s planning mentions 2027.

Were SMEs hit harder than the allowances parents?

If the SME directorate accounted for 89 percent of RAM usage, the obvious question is whether small and medium-sized business was hit harder than the parents in the allowances scandal. The honest answer has three layers: scale, depth and redress.

Scale: yes. The 2017 logging shows the SME directorate ran 19,091 RAM queries that year, against 1,318 by Direct Taxes, 197 by Large Businesses, 137 by Customs and 136 by the FIOD. The system itself held 2.2 million entities, including the records of 1.9 million businesses, against an SME segment of roughly 1.5 million companies. Over twenty years the system produced an estimated half million selections, at an average of more than 20,000 per year. Our earlier investigation conservatively estimated 100,000 to 300,000 unique affected entrepreneurs. By comparison: the allowances scandal started at a count of 26,000 families, but the redress registration of the Tax and Customs Administration’s Benefits department stood at 44,102 recognised affected parents in September 2026, out of 69,722 registrations, and the FSV logged some 180,000 registrations. In sheer numbers the SME selections remain a multiple of the group of allowances parents and at least the size of the FSV. Small and medium-sized business was not a side branch but the main channel of the system: its largest user and, through the SME intelligence staff, also its builder.

Depth per family: no, with a caveat. The allowances blow was immediate and existential: housing and healthcare allowances cut off, recoveries, mounting debts, eviction. And the count of children grew with every study. The Van Rhee commission handled 1,115 reports. Statistics Netherlands counted 2,090 children placed outside the home between 2015 and mid-2022 at the end of 2022, with 645 of those placements still running at the measuring point. The data link under the Temporary Act on the Exchange of Personal Data ultimately identified 4,791 unique children from nearly 3,000 families and 2,888 parents who faced an out-of-home placement over 2023 through 2026. Marriages and families broke; suicides occurred among the victims. The SME route looks milder on paper: return, selection, audit, additional assessment. That impression does not survive the composition of the segment. The 2026 Annual Plan counts 1,031,000 micro-entrepreneurs, self-employed included. They have no limited company shield: a sole proprietor or partnership attacked is personally liable, and the wealth comparison for which ITE is building the new VVG tool reaches into the household books and the partner’s assets. Additional assessments run over multiple years, penalties can reach 100 percent, and where an actual employment relationship is suspected the reassessment period extends to twelve years. Anyone reported for criminal investigation under the AAFD protocol usually loses the business outright, and with it income, home and pension. For a self-employed entrepreneur that is the same existential harm as for an allowances parent, only by a slower route and without public recognition as a victim group.

Redress: here lies the real inequality.

For the SME group only the Selection at the Gate compensation scheme exists, and it covers solely blocked income tax returns over 2013 through 2017. Anyone selected through RAM, the FSV or the tips route falls outside it. The state secretary excluded research into third-party effects in March 2025, and declared in December 2025 that there was “no reason” to assume citizens had been harmed. What distinguishes the SME group is therefore not the depth of individual suffering but the combination of the largest scale and the complete absence of redress. Equal unlawfulness, unequal legal consequences: precisely the test that Article 1 of the Constitution forbids.

The redress itself creaked too, according to the watchdogs. The National Ombudsman has followed the UHT’s complaints handling since late 2020. His second monitor (October 2021) reported that the complaints team was “not prepared for that situation” when complaint numbers surged — overwhelmingly about processing times — and that statutory deadlines were exceeded in more than a quarter of cases; the July 2022 report “Waiting for redress, how much longer?” repeated the criticism. The UHT’s own satisfaction survey now shows 84 percent of parents feel “heard” — but on a sample of 49 participants in the second quarter of 2026, where a year earlier 46 percent was measured on 143 participants. The Dutch Data Protection Authority received more than 13,500 complaints and tips in 2025, 75 percent more than in 2024. And in February 2026 five local youth ombudsmen bundled more than 1,800 reports from young people in allowances families — 55 percent took longer over their education, many carrying study debts — again without a generic scheme following. OpenBrief’s own anonymous reporting portal, BAM (Report Anonymously), counted 209 reports from victims in October 2026: 45 about the UHT, 27 about the OTB, 24 directly about the Belastingdienst, and 83 from victims abroad, the group waiting longest for redress.

The disproportion repeats. Of the victims in the allowances scandal, 71 percent had a migration background. In RAM, 11,236 citizen service numbers were selected on nationality between 2012 and 2014, and among the 1,170 extracts found were 14 spreadsheets with selection on nationality and 35 with selection on postcode. The same mechanisms, in a larger population.

This is not a competition in suffering; pain cannot be averaged. But the question admits a twofold answer. By scale and by redress, small and medium-sized business is worse off than the allowances parents. By immediate individual impact, the allowances blow was harder and faster. Together the groups form one system with one striking ratio: 89 percent of the usage, and a redress scheme that applies to zero percent of the SME victims. Whoever called the allowances scandal the tip of the iceberg meant this main channel.

VVG and FAM: profiling, rebuilt

The most current evidence of the continuing line sits under a heading “Models” in the SME work and production plan 2025-2, the plan for the second half of 2025. Full passage:

In cooperation between the ITE teams and DF&A, models are being built to support the ITE work process. These concern a wealth comparison tool (VVG) and the Fiscal Analysis Model (FAM). For both models the application can be broader than ITE.

For the VVG, a pilot ran in 2024, and in 2025 it is being examined whether and how the VVG can be taken further into production and for which process the VVG can offer support (signal model, selection, assessment, audit).

For the FAM, the initial focus was on developing a dashboard function. To test the usability of the dashboard, a three-month pilot started in July 2025. The FAM aims to offer support at a strategic level with insight into sectors/target groups and can thereby serve as underpinning for choices at a strategic level. In the FAM pilot, an enforcement adviser from the client treatment department is participating on the theme of “labour relations enforcement”.

Two models, two functions. The VVG is a wealth comparison tool: an instrument that sets an entrepreneur’s declared assets against spending patterns, the classic opening of any audit into unexplained wealth. The policy question left open in 2025 is which process the tool will feed: signal model, selection, assessment or audit. Those are the four steps of the selection process RAM once automated. The FAM is more strategic: a dashboard mapping sectors and target groups to underpin “choices at a strategic level”, with labour relations enforcement as the first pilot theme, the domain of sham self-employment and the labour misconduct of the LSN and AMU approaches. The intention for end 2025 was a Minimum Viable Product; in the 2026 plan the pilot has run and production is scheduled.

A legal mirror is available. The Data Protection Authority ruled in July 2025 that successors of RAM with comparable risks had to be phased out immediately. That demand does not apply to VVG and FAM, because they are new and not yet in production. But the question the KPMG investigation put to RAM deserves an honest answer here too: is there a legal basis, is there a DPIA, are the selection criteria recorded traceably, and does the entrepreneur who gets selected know that a model co-decided? The 2026 work plan promises for DF&A’s signal models a “signal administration” registering treatment advice so that criteria are “recorded in an automated and traceable manner”. That sentence is in itself an admission of what used to be missing.

For the Risky Networks team’s ARN Dashboard, a DPIA plus a Privacy Risk Assessment lay with the data protection officer in April 2025, before the pilot began. That is the right order. Whether VVG and FAM development underwent the same test appears in no public document.

What this means for the entrepreneur

For an entrepreneur approached by ITE, the legal position does not change on paper, but the practice does. An ITE investigation is an audit under the General Act on State Taxes, with the same rights and the same duty of reasoning as any other. Still, the context differs in essence. The trigger more often lies in a signal from outside the tax system, a tip, a RIEC weighing decision, a labour inspection, than in a deviation in the return. The investigators sit in one team with assessment, collection and the penalty function. And according to the organisation’s own work plans, the outcome can proceed to the AAFD protocol for criminal referral.

Anyone wanting to know how a file came about has three routes. Article 5.5, first paragraph, of the Woo grants access to one’s own file, including internal signals. Articles 15 and 22 of the GDPR grant access to processed personal data and the right to human review of automated decisions; Article 22 does not automatically apply to every risk selection, but the criteria must be public or obtainable. And Article 3:46 of the General Administrative Law Act requires that an additional assessment or penalty carries a factual and legal motivation: which signals, which selection, which investigation. A motivation that stops at “following information received” no longer passes, certainly now that the work plans show how many kinds of signals exist.

The older scandals open a fourth route. For selections that wrongly led to extra burdens, the compensation scheme for Selection at the Gate has existed since 2023. For RAM-selected entrepreneurs that scheme was not written, and the state secretary excluded research into third-party effects in March 2025. Whether ITE files trace back to old RAM or FSV selections is therefore not answered but precisely left open.

Open questions

This investigation rests on public documents. What is in them justifies the conclusion that EOS merged into ITE and that the profiling and undermining lines converge there. What is not in them is at least as relevant.

How many audits did the ITE teams conduct in recent years, and against whom? The production plans state hours, not numbers. Which selection criteria feed the 12 reporting desks now that the FMI application routes on BSN and RSIN? Do the signals in the RIEC pre-weighing still contain nationality or postcode markers, as the sweep operation and the parliamentary letter on RAM extracts found? Have DPIAs been drawn up for VVG and FAM, and where can they be obtained? And how many ITE files reach back to selections from the RAM period can no longer be reconstructed for lack of searchable logging, as KPMG acknowledged.

The Belastingdienst can lay all of this open with a Woo decision. Until then, what a participant in the July 2022 internal session noted about intensive supervision holds, meant as praise and now readable as a warning: the auditor who finds a case should not lose it, but “become part of the investigation group within ITE or get involved in the criminal investigation”. That is the machine, described by the people operating it.

Frequently asked questions

What is EOS at the Belastingdienst?

EOS stands for External Government Cooperation, the team that ran the tips-and-informants mailbox and, per the Fraud Enforcement Plan, supplied pre-weighing documents to the RIEC.

What is ITE at the Belastingdienst?

ITE stands for Intensive Supervision and External Government Cooperation: eleven cross-SME teams with audit, assessment, collection, an information desk and an AWR liaison officer, totalling 562,200 hours in 2026.

When did EOS become ITE?

Between 2021 and 2025: after tips processing was halted on 2 July 2020, the Belastingdienst built the reporting landscape that since 2026 routes on BSN and RSIN to twelve desks.

What role does the WGS play for ITE?

The Data Processing by Partnerships Act (in force 1 March 2025) gave the RIEC, LIEC and iCOV cooperation in which ITE works their statutory basis, years after the practice was running.

How many SME entrepreneurs were affected by profiling?

Conservative estimates run to 100,000 to 300,000 affected entrepreneurs; SMEs ran 89 percent of all RAM queries and held their own main purpose authorisation on the FSV.

References

1. Work and Production Plan SME 2025-2, Belastingdienst, open.overheid.nl
2. Work and Production Plan SME 2026, Belastingdienst, open.overheid.nl
3. Woo decision tips and informants (partial decision 2), Belastingdienst/SME, 6 March 2023, open.overheid.nl
4. Progress report Sweep Operation Lists (Prio 1b), 18 February 2022, rijksoverheid.nl
5. Prio 1b: Appendix final overview of lists, 18 February 2022, rijksoverheid.nl
6. Parliamentary letter Fraud Signalling Facility and institutional racism, 30 May 2022, rijksoverheid.nl
7. Background document Intensive Supervision session, 8 July 2022, Childcare Allowance Information Point
8. Vacancy Supervisory Officer ITE, werken.belastingdienst.nl
9. Parliamentary letter policy response RAM, 6 March 2025, rijksoverheid.nl
10. DPA letter investigation RAM and similar systems, 9 July 2025, reference 2025-002145
11. Compensation for Selection at the Gate Act, 19 December 2023
12. Data Processing by Partnerships Act, in force 1 March 2025
13. RIEC and undermining: contact structure with ITE team leader, Hillegom municipal council
14. Open Brief: RAM and SME profiling, the tip of the iceberg
15. Open Brief: RDM, the Regional Data Model
16. Open Brief: RAM never disappeared
17. Open Brief: RIEC/LIEC and the Belastingdienst
18. Open Brief: The algorithm affair of the Netherlands

This investigation is based solely on public sources. Corrections and additions are welcome via the contact page.

Special thanks to Mourat for supplying documents. You’re a legend.