The use of Artificial Intelligence by tax authorities: advantages, risks and challenges
Magda Purice
Posted On September 30, 2024
Authors: Andra Cașu, Partner, head of the Direct Taxes department, EY Romania, Teona Braia, Manager, Direct Taxes, EY Romania
Tax authorities around the world have begun to analyze the opportunity and even implement various solutions based on Artificial Intelligence (AI) and machine learning to improve the efficiency of administrative processes, detect tax evasion and optimize interaction with taxpayers. Regulators, for their part, are working hard on the framework of rules by which new technologies are designed and used correctly, transparently and non-discriminatory. We analyze, in this context, how the tax authorities have already implemented AI-based solutions, but also the current trends in this endeavor.
The tax authority in Poland can take measures, such as blocking an entrepreneur’s bank account or canceling his VAT registration, depending on the risk indicator provided by the Compensation System (STIR), the mechanism that establishes a risk indicator for each taxpayer in part, calculated with the help of some AI algorithms. In Ireland, AI has been used to automate the process of categorizing and routing taxpayer inquiries to the appropriate subject matter experts. This resulted in a significant increase in accuracy levels from 70% to 97% and reduced total targeting time by over 24 hours. In Israel, the tax authority is using AI in an automated property tax assessment project. This project combined AI components that simulate the work of a tax inspector with the computing capabilities of online systems, reducing assessment time and allowing officials to focus on more complex cases.
In France, the French tax authority uses AI to improve the efficiency of their legal function. Basically, AI is used to provide legal staff with proposed answers to simple and recurring requests, detect new trends earlier, and even predict future legal developments. In Sweden, the tax authority uses AI to interpret and convert handwritten text into digital text (n.b. in Sweden, individuals and legal entities can also submit tax returns on paper. Every year, the authority in Sweden receives approximately 150,000 paper forms containing handwritten information). In addition, AI is used to distinguish high-risk from low-risk cases, allowing operators to identify and reject incorrect tax deduction claims. The AI model is trained on data from previous investigations, allowing it to detect common characteristics of those who do not respond to substantiation requests.
Trends in the Adoption of AI Solutions
Development and Ethics in Use We see that more and more tax authorities are increasingly using AI to offer new services to taxpayers and to streamline their internal processes. But there are also challenges and there are many aspects that need to be taken into account, including the constant concern related to the ethical use of artificial intelligence solutions, especially in such a complex field as taxation. Authorities must ensure that AI systems are transparent, respect taxpayers’ rights and do not operate with unwanted discrimination.
That’s why both regulatory authorities and GDPR and cybersecurity experts recommend testing AI solutions in controlled environments, pre- and post-implementation audits, and performing impact assessments to prevent risks. A good practice comes from Canada which has experienced the implementation of AI solutions by the Canadian Revenue Authority (CRA). To ensure the responsible use of AI, the CRA has established a governance structure, which includes the Artificial Intelligence Directive implemented in January 2021.
This outlines roles and responsibilities within the CRA and is supported by the mandatory use of the Impact Assessment and Alignment Tool Algorithmic (AIAA). The AIAA tool is designed to evolve as AI governance matures and the AI context changes rapidly, supporting informed oversight and promoting transparency.
The use of AI by tax authorities has the potential to radically transform the way taxes and duties are administered in the long term, bringing significant benefits in terms of efficiency and fairness. Without a doubt, we expect that in Romania, the use of AI in matters related to taxation will continue on a similar trend. It will be the natural course of things, therefore a prominent question mark will remain that related to a proper administration of AI in the area of regulation, ethics and transparency, without generating discrimination at the level of taxpayers. We already see that technology has taken a strong shape in Romania, especially in the area of indirect taxes.
But given the complexity and diversity of international tax regulations, as well as their incidence at the local level, it is expected that in the coming period the implementation of technology and AI in taxation will continue on an upward trajectory.
Trending Now
