The use of machine-learning algorithms by tax administrations of EU Member States is not more than a decade old. Yet, despite being at a relatively early stage of development and use, some machine-learning systems have already been the subject of landmark court litigations, which will undoubtedly shape the future of tax algorithmic governance for the years to come.

The following section describe the important facts and key takeaways which can be inferred from these landmark rulings.

In February 2020, the Court of the Hague halted the use of Systeem Risico Indicatie, a risk-scoring algorithm to be used by the Dutch tax administration following the claims by civil society interest groups that the algorithm would violate several fundamental rights of taxpayers residing in The Netherlands.

On the 15th of January 2021, the entire cabinet of Mark Rutte, Prime Minister of the Netherlands was forced to resigned amidst the revelation that the social security unit of the Dutch tax administration had unlawfully discontinued and requested the reimbursement of childcare allowances to 35.000 parents, leading those into serious financial hardship. A machine-learning algorithm used by the tax administration discriminated thousands of welfare recipients.

The 17th December 2021, the Slovak Supreme Constitutional Court ruled that a State-mandated electronic cash register system which obliged entrepreneurs to transfer VAT data to the Slovak Financial Directorate in real-time, was incompatible with taxpayers’ fundamental rights.