AI tools have the potential to profoundly reshape the way the taxation field operates. Governments are progressively leveraging AI tools to streamline compliance procedures, tax audits, and administrative claims. Consequently, the deployment of this technology can yield measurable gains in revenue recovery and streamlining back-office tasks, freeing tax authorities’ employees to do more complex work. Conversely, taxpayers can benefit from faster refunds, fewer bureaucratic hurdles, and more personalised guidance.
However, this technological leap comes with its own risks and ethical quandaries. The implementation of AI's predictive models has the capacity to transform the human and transparent tax administration decision-making process into an opaque one driven by inscrutable algorithms. Systems trained on massive data sets, by ignoring nuanced cases or because of embedded biases, may prove to be extremely dangerous. Thus, the efficiency allure of AI-based solutions must be balanced by safeguarding fundamental legal rights and guarantees of the taxpayers.
While paper-based tax procedures worldwide have shifted to digital platforms, and repetitive tasks are increasingly handled without human input, AI advances introduce a significant qualitative leap. AI tools have the capability to create profiles, assign risk scores, and even predict behaviours, which may allow tax authorities to provide tailor-made assistance to taxpayers but also enable the selection of individual taxpayers for audits or to assess the tax or penalties they must pay, based on probabilistic outcomes. Therefore, algorithmic solutions have the capacity to expand the toolbox of tax authorities, but they also pose a challenge on the right to reasoned decisions, and they risk reinforcing societal biases.
Furthermore, as AI-powered technologies become capable decision-makers (rather than being mere assistants), there is an ever-growing number of questions regarding the accountability of such decisions:
Who should be held legally responsible for erroneous decisions (the entity that designs an algorithm, the tax authority that implements the algorithm, or the algorithm itself)?
Under what terms may they be appealed?
How can aggrieved taxpayers seek reparations due to decisions that might not be fully understandable by anyone?
Consequently, the implementation of AI-enabled automated decision-making by the tax authorities must be designed and governed with clarity in mind, so that every taxpayer subject to an adverse ruling has the capacity to understand the reason for the decisions, and to question them by appealing to qualified human authority. Governments should prioritise the commitment of preserving and reinforcing hard‑won guarantees, as well as respecting principles of transparency, accountability, and justice over blindingly being on the frontier of this technology.
Ensuring accountability in automated tax decisions
The essential procedural safeguards should include the following:
The prohibition of fully automated binding tax rulings, and the requirement of human intervention at every pivotal stage of the tax authorities’ decision-making process. AI tools should enhance tax administration but cannot replace human judgement.
Access to sensitive personal data should be strictly regulated, grounded in the principles of proportionality, requiring explicit consent from taxpayers or a judicial warrant.
Taxpayers should be notified whenever AI systems are involved in the processing of their cases, including the criteria and purposes behind any automated action.
Specific compensation mechanisms for damages caused by AI-induced errors or injustices should be established.
Information campaigns should be set up whenever an AI feature is implemented within the tax authorities’ workstream, to promote transparency and demystify the way tax authorities leverage AI-powered solutions in their decision-making process.
Final thoughts on safeguarding taxpayer rights in the AI era
The momentum of the deployment of AI systems by tax authorities seems to be unstoppable. Because of this, profound opportunities and challenges lie ahead. The success of this endeavour will depend not only on future technological improvements but also on the willingness to engage critically with this technology to forge a delicate balance that ensures AI is harnessed for the genuine public good, fortifying (rather than violating) the legal rights and guarantees of taxpayers.
