State of use of AI tax systems
The Belgian Tax Administration (SPF Finances) has been using machine-learning algorithms since as early as 2014 according to official documents from IOTA. Incidental mentions report that some models may have been used tested and used already earlier than 2014, yet these sources are not corroborated by official testimonies of the SPF.
What functions are performed with AI?
The machine-learning algorithms of the Belgian Tax Administration perform a wide range of functions:
1. AI web-scraping: the SPF Finances makes use of an algorithm to automatically collect taxpayer data from e-commerce and e-sharing platforms, e.g. Amazon, Airbnb, eBay, 2emeMain, etc.
2. Social Network Analysis (SNA): the SNA algorithm 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/or qualitatively measure relations between the nodes.
3. Internal risk-management: The SPF Finances uses a suite of algorithms to predict the risk that taxpayers do not pay their taxes due, following a letter from the bailiff (‘Pegasus’), or following a call from the outbound call center (‘Iris’). These algorithms assist the SPF with their internal case management, and predict what course of action is most appropriate for the administration, based on historical taxpayer data, e.g. for taxpayers who are notoriously compliant/non-compliant these models prescribe a more coercive/cooperative course of action, and vice versa.
4. External risk-management (risk-scoring algorithms): The SPF Finances uses a suite of algorithms to predict specific risks of non-compliance of individual taxpayers, ‘Hermes’ predicts the risk of bankruptcy within a 12 months period for legal and self-employed persons, ‘Delphi’ predicts the solvency rate for natural, legal and self-employed persons. The SPF Finances also uses models to segment taxpayers into categories of risks to develop their annual audit plans, and select taxpayers with high-risks of non-compliance for further audits by human tax officials.
5. Nudging: the Belgian Tax Administration uses an algorithm to adapt the language of standard communication of taxpayers based on an analysis of individual taxpayer data.
What data can be processed by these systems?
The data used for these models is not specified. In accordance with the ‘Loi du 3 août 2012 portant dispositions relatives aux traitements de données à caractère personnel réalisés par le Service public fédéral Finances dans le cadre de ses missions‘ all taxpayer data is stored in a central data warehouse, and can be used for the development or use of algorithms.
Art. 327 of the Belgian Income Tax Code establishes that other governmental institutions must provide to tax officials any information that would be deemed relevant for the collection and levy of taxes. However, this information cannot be communicated or exchanged without the ‘express authorisation’ of the Federal Public Service. This express authorisation requirement was adapted by Art. 72 of the law of 5 September 2018, ‘law instituting the information security committee’ which provides that information can only be exchanged in accordance with protocols drawn up by federal authorities. (De Raedt, 2021)
Are these systems regulated by specific norms?
There are no specific legal norms which regulate the use of machine-learning algorithms. These algorithms have been implemented on the basis of managerial decisions of the SPF Finances. The norm which regulates the use of taxpayer data by the SPF does not make explicit mentions of the exact machine-learning algorithms, with the exception of risk-scoring, see: ‘Loi du 3 août 2012 portant dispositions relatives aux traitements de données à caractère personnel réalisés par le Service public fédéral Finances dans le cadre de ses missions, Art. 5’.
References:
Legislation:
- Article 327-337, Belgian Income Tax Code
- Art. 3- 7, Loi du 3 Août 2012 portant dispositions relatives aux traitements de données à caractère personnel réalisés par le Service public fédéral Finances dans le cadre de ses missions’
- Loi du 30 Juillet 2018, relative à la protection des personnes physiques à l’égard des traitements de données à caractère personnel
- Art. 72, Loi du 5 Septembre 2018 instituant le comité de sécurité de l’information
Doctrine:
- OECD, Tax Administration Series 2017, pp. 110-111 (Box. 6.15).
- OECD, Tax Administration Series 2019 Comparative Information on OECD and other Advanced and Emerging Economies, p. 52.
- OECD Forum on Tax Administration (FAT), Successful Tax Debt Management (OECD, 2019), p. 21 & p. 68.
- IOTA Papers, ‘Nudging in the Context of Taxation’ (February 2019), p. 7
- D. Dierickx ‘The Belgian compliance model and the methodology to obtain data from “Sharing Economy” platforms’ (IOTA 2017)
- D. van Hout, ‘Gedragsbeïnvloeding in het belastingrecht: Are you ‘nudge’. Tijdschrift voor Fiscaal Recht, 2018 [549-550], p. 928-936.
- S. De Raedt, De draagwijdte van het recht op privéleven bij de informatie-inzameling door de fiscale administratie, Brussel, Larcier, 2017, 1-3
- S. De Raedt, ‘De privacybescherming van de belastingplichtige bij de digitale informatie-uitwisseling. Of hoe de GDPR een stap achteruit kan zijn’ (Peeters et al. eds.) Liber Amicorum Stefaan Van Crombrugge. (Roularta, 2020): http://hdl.handle.net/1854/LU-8698136
- S. De Raedt, ‘Privacy concerns related to the digital tax administration 2.0′ in (1st ed. Hervé Jacquemin eds., Larcier 2021) Time to Reshape the Digital Society – 40th Anniversary of the CRIDS
- LaLibre, 16 May 2011: https://www.lalibre.be/economie/entreprises-startup/2011/05/16/le-bisc-nouvel-outil-charge-de-lutter-contre-la-fraude-sur-internet-UXFT5ZKR5FH5RAHXIACSLW3TEE/ - last accessed June 2022.
- See also lecture of Van Vlasselaer et al. (KUL) – https://limo.libis.be/primo-explore/fulldisplay?docid=LIRIAS1834677&context=L&vid=Lirias&search_scope=Lirias&tab=default_tab&lang=en_US&fromSitemap=1 - last accessed June 2022.