State of use of AI tax systems
The Netherlands has, with Member State such as Belgium, Denmark or Sweden, always been at the forefront of automation and the use of technology to combat tax fraud and ensure tax compliance. As early 2004, the Netherlands had developed an AI web-scraping tool ‘XENON’. In 2006, the Netherlands was one of the Member States piloting the EU Fiscalis Risk Analysis Project Group, drafting guides for other EU Member States’ tax administration on how to manage risks to tax compliance.
Accordingly, it is not surprising that in 2007, the Netherlands was one of the leading Member States experimenting with data mining and machine-learning to track tax evaders, in particular those operating through digital ecosystems on the internet.
What functions are performed with AI?
As of February 2022, tax machine-learning algorithms are performing four types of functions in the Netherlands:
- AI web-scraping ‘XENON’: Web-scraping, sometimes also referred as ‘web-crawling’ or ‘web-spiders’, are algorithms that scout and surveil webpages and links on those page, automatically collecting data from these pages, these social media or e-commerce platforms and match it, through data matching, with data already present in the databases of the tax administration. This process does not necessarily require artificial intelligence, but can be perfomed with robot process automation. However, machine-learning is believed to be more sophisticated, in particular in the matching phase, as it can more accurately match the data collected with the data already retained by the administration. Sources indicate that Austria and Denmark are also using such system regarding their taxpayers’ webpages and data on the internet.
- Social Network Analysis (SNA): the SNA system visually represents a network of individual taxpayers using graph theory, it represents a network of taxpayers as a combination of nodes for individuals or points of interests, and lines which quantitatively and qualitatively measure relations between the nodes.
- External risk-management (risk-scoring) ‘SyRI’: Systeem Risico Indicatie or SyRI is a machine-learning system which predicts the risks of tax fraud or tax non-compliance associated with individual taxpayers, segmenting taxpayers into categories of risks to subsequently select taxpayers for further audits by tax officials. The use of this system was halted by the Court of the Hague in 2020, after it found that Art. 65 of the SUWI wet, the norm which regulated the use of the tool lacked transparency and therefore breached Art. 8(2) of the ECHR. (see more on that judgment in the publication section). The Dutch legislator is currently in the process of adopting a new set of provisions to bring SyRI back to the table, a project which academics have qualified as a ‘Super SyRI’.
- Nudging: nudging tools are algorithms which adapt the language used in standard communication to taxpayers, based on behavioural insights derived from scientific literature and inference made from individual profiling of historical data on individual taxpayers. Reportedly, one of these nudging tools is targeting recent divorcees, as these taxpayers have been found to be a particular group prone to tax non-compliance.
What data can be processed by these systems?
The specific data used for the development and use of these tax machine-learning algorithms is not specified. However, Art. 64 (1) (a-d) of the SUWI wet provides that any goverment records or data held by Dutch administrative bodies can be used for the purpose of preventing or combatting tax fraud. This literal ‘carte-blanche’ regime of data use was one of the reason for the judgment in SyRI and for the decision of the Court of the Hague to temporarily halt the use of SyRI by the Dutch tax administration.
Are these systems regulated by specific norms?
Among all the tax machine-learning algorithms aforementioned only SyRI was regulated by specific legislative norms, namely Art. 64 & 65 of the SUWI wet. However, as Art. 65 of the SUWI wet was found to be in breach of Art. 8(2) of the ECHR, these algorithms are currently not regulated by specific legal norms. That is, until the Dutch government adopts its so-called ‘Super SyRI’ legislation, which will regulate the use of its external risk-management (risk-scoring) algorithm.
- Wet van 29 november 2001, houdende regels tot vaststelling van een structuur voor de uitvoering van taken met betrekking tot de arbeidsvoorziening en socialeverzekeringswetten (Wet structuur uitvoeringsorganisatie werk en inkomen – SUWI wet), Article 64 & 65.
- Court of the Hague, Systeem Risico Indicatie (SyRI) case (English translation): https://uitspraken.rechtspraak.nl/inziendocument?id=ECLI:NL:RBDHA:2020:1878
- European Commission (DG TAXUD), Risk Management Guide for Tax Administration – Fiscalis Risk Analysis Project Group (February 2006), p. 67 – available at: https://ec.europa.eu/taxation_customs/system/files/2016-09/risk_management_guide_for_tax_administrations_en.pdf
- OECD, Advanced Analytics for Better Tax Administration, p. 21.
- Belastingdienst, Bijlage bij besluit met kenmerk 2018 – 000018765 (Documenten 1 – 39) – 20 May 2019 – available at : https://www.rijksoverheid.nl/binaries/rijksoverheid/documenten/wob-verzoeken/2019/05/20/wob-besluit-over-monitoring-sociale-media-bij-de-belastingdienst/4+Bijlage+4+bij+Wob-besluit.pdf
- Wired (25 January 2007), ‘Tax Takers Send in the Spiders’ – available at:https://www.wired.com/2007/01/tax-takers-send-in-the-spiders/
- J. Liu, Y. H. Tan and J. Hulstijn, ‘ IT enabled risk management for taxation and customs’ in M. Wimmer (eds.), Electronic Government: 8th Conference Proceedings (Springer, 2009), p. 384.
- IOTA Papers, Nudging in the Context of Taxation (February 2019), p. 7
- D. van Hout, Gedragsbeïnvloeding in het belastingrecht: Are you ‘nudge’. Tijdschrift voor Fiscaal Recht, 2018 [549-550], p. 928-936.
- Viktoria Wöhrer, ‘Effective Taxation Versus Effective Data Protection’ & Tina Ehrke-Rabel, ‘Big Data in Tax Collection and Enforcement’ in Haslehner et al. (eds.) Tax and the Digital Economy: Challenges and Proposals for Reform (Kluwer Law, 2019), Chapter 11 & Chapter 13