Proceedings of the International scientific and practical conference ―Science, Technology and Culture: Interaction, Evolution and Progress‖ (December 21-23, 2025) / Publisher website: www.naukainfo.com. – Copenhagen, Denmark, 2026. – 161 p.

35 violations without delays. [3, p. 2-3]. This helps to avoid fines because of random mistakes, which is extremely important for small and medium businesses. The usage of RegTech is highly beneficial for state tax administrations. The analysis of big data and machine learning enables automated detection of suspicious operations and deviations in reports. For example, the Australian tax service has found 530 million Australian dollars of unpaid taxes with its help [4]. Practical cases of big consulting companies demonstrate the effectiveness of implementing AI in tax governance. Deloitte after integrating Zora AI has reduced tax-related process costs by 25% and improved the productivity of specialists by 40%. This was possible because of reports automation and workflows. On the other hand, EY developed its own platform based on NVIDIA agent models. This platform annually supports about 80,000 specialists and processes millions of compliance operations, ranging from preparing tax returns to assessing risks [5, p. 110]. These examples illustrate how AI frees specialists from routine, enabling them to work on complex tasks. Despite all the described benefits, using AI models for compliance operations imposes two major risks. First of all, poor security controls are leading to the leaks of confidential information. Secondly, AI systems, trained on inaccurate or incomplete tax data, spread biased models, such as marking particular groups or branches as erroneously classified as high-risk due to biases in training datasets [6]. There are several ideas to overcome these challenges. In order to protect the tax payers data researchers propose implementing a system of data anonymisation and end-to-end encryption. In addition to this, regular algorithm audits and checking sources of training data are essential elements for keeping the correctness of the advice from AI. To enhance the effectiveness of such safeguards, it is important to ensure transparency and control in using AI at preparing tax reporting [5, 126]. To conclude, AI is becoming an integral part of tax compliance. This technology lowers operational expenses and increases the accuracy of making decisions. However, despite declared efficiency, AI models remain ―black boxes‖ and make regulatory audit complex. This threatens inaccurate decisions in the tax