The Application of Generative Artificial Intelligence in the Analysis of Tax Representations for Criminal Purposes in the Federal Public Prosecutor’s Office: Reflections on Procedural Efficiency

Frankslany Silva dos Santos Reis; Mateus Otoni Silva

DOI: 10.22167/2675-6528-202602923

Article derived from a Course Conclusion Work (TCC), with content based on the student’s original work and adapted to the editorial format of the E&S Magazine with the support of the ResumeAI tool, an artificial intelligence solution developed by Instituto Pecege for textual synthesis and organization.

Summary

Digital transformation in the Brazilian public sector has driven the incorporation of Generative Artificial Intelligence as an institutional support tool. The study analyzed the application of Generative Artificial Intelligence in the Federal Public Prosecutor’s Office, focusing on the efficiency of tax processes and the analysis of Fiscal Representations for Criminal Purposes. The objective was to examine how the technology contributed to the analysis of large volumes of data, the cross-referencing of information, and the drafting of legal documents. The research, quantitative and applied in nature, with exploratory and descriptive objectives, used bibliographical, documentary, and field research through structured questionnaires. The sample included four members and three interns from the institution. The results indicated that Generative Artificial Intelligence promoted gains in speed in procedural analysis, standardized activities, and assisted in the systematization of complex information, directing focus towards analytical activities. However, the need for human supervision to validate the produced information was verified, given the requirement for legal accuracy. It was concluded that the use of Generative Artificial Intelligence represented a significant advance in institutional efficiency, although it still requires attention to ethical, transparency, and data protection aspects.

Keywords: Procedural efficiency; Generative Artificial Intelligence; Federal Public Prosecutor’s Office; Tax Representations.

1. Introduction

Artificial Intelligence (AI), although present for over two decades, has gained significant visibility with the popularization of systems based on language models, such as ChatGPT. This advancement has intensified debates related to privacy, information security, and the ethical use of personal data, provoking significant transformations in various sectors of society, including the justice system (Stryker and Kavlakoglu, 2024).

AI-based technologies can bring benefits to different professional fields, assisting in the diagnosis and prediction of diseases, in the elaboration of legal petitions, in the prediction of sentences, in the analysis and proposition of network improvements, in information and reference, and in the descriptive treatment of resources (Stryker and Kavlakoglu, 2024). UNESCO (2021) highlights that artificial intelligence tools, especially generative ones, have the capacity to process large volumes of data, respond to complex demands, and assist in content production, being increasingly used in academic and professional contexts.

In the Brazilian context, digital transformation in the public sector intensified from 2020 onwards, with the federal government establishing the integral digitalization of public services as a goal. Law No. 14.129, of March 29, 2021, formalized this directive, providing for the principles, rules, and instruments of Digital Government, with a focus on innovation, administrative efficiency, and expanded access to public services. This legislation encourages the use of emerging technologies as a means of modernizing public administration (Brazil, 2021).

In practice, Artificial Intelligence can be employed in information triage, in the cross-referencing of data from different control bodies, and in the drafting of opinions and reports. This contributes to greater speed, standardization, and accuracy in the core activities of Public Administration. Costa (2020) points out that the incorporation of these technologies represents a relevant opportunity to reconfigure the evaluation of the Judiciary’s efficiency, given the gains observed in the agility of legal work.

Within the scope of the Federal Public Prosecutor’s Office (MPF), the adoption of Artificial Intelligence occurred gradually and in a structured manner. The first initiatives, developed between 2020 and 2021, focused on internal automation and case screening projects, with the predominant use of predictive AI models aimed at identifying patterns and classifying documents. In 2022, the MPF consolidated the use of artificial intelligence through the TRIA System, automating the screening of habeas corpus and police investigations, with reduced analysis time and greater efficiency in case prioritization (MPF, 2023).

In 2023, the MPF advanced to the application of Generative Artificial Intelligence in a pilot project aimed at fiscal representations for criminal purposes (RFFP), enabling the extraction of information from complex documents and the automated generation of drafts in investigations involving crimes such as tax evasion, smuggling, illicit import, and money laundering. Given this scenario, the following research problem arises: to what extent does the use of Generative Artificial Intelligence within the Federal Public Prosecutor’s Office contribute to institutional efficiency and speed in the analysis of Fiscal Representations for Criminal Purposes (RFFP), especially when compared to the traditional processing model?

The relevance of the topic is justified by the continuous search for greater efficiency and speed in public administration, especially in complex processes such as the analysis of RFFP, where Generative Artificial Intelligence can optimize work and support decision-making. The present work aimed to analyze the use of Generative Artificial Intelligence within the Federal Public Prosecutor’s Office, with an emphasis on its application in the analysis of Fiscal Representations for Criminal Purposes (RFFP) as an instrument to support institutional efficiency and procedural speed.

2. Material and Methods

The methodology adopted in this work was characterized as quantitative research, of an applied nature, with exploratory and descriptive objectives, according to the classification by Chizzotti (2003). This approach enabled analysis through the collection of structured empirical data, allowing for the measurement and comparison of participants’ perceptions regarding the impacts of implementing Generative Artificial Intelligence tools on internal process efficiency, aligning with the methodological orientation of Martins and Theophilo (2009).

Regarding the technical procedures employed, the research utilized a combination of methods, including bibliographic research, documentary research, and field survey. Such procedures, according to Gil (2017), are suitable for the in-depth analysis of social and institutional phenomena, such as the object of this study.

The bibliographic research was conducted based on a survey of specialized literature. Books, scientific articles, technical reports, and institutional publications addressing topics such as Artificial Intelligence, digital governance, and innovation in the public sector were consulted, providing the necessary theoretical foundation for understanding the phenomenon investigated.

Complementarily, the documentary research consisted of the analysis of primary and secondary sources. The current legislation, institutional regulations, and public documents related to the adoption of Artificial Intelligence technologies within the Federal Public Prosecutor’s Office were examined. This procedure allowed for an understanding of the legal and institutional framework that governs the implementation of these tools.

The field survey was conducted by applying structured questionnaires, anonymously and voluntarily. The participants were individuals linked to the Federal Public Prosecutor’s Office (MPF) who directly work in data analysis, case management, and the production of legal documents, areas directly impacted by Generative Artificial Intelligence.

The universe selected for participation in the research was composed of seven people. The final sample brought together four members of the MPF and three interns from the institution, totaling seven respondents. MPF servers were not included in the universe for the application of the questionnaire, not being represented in the results of the field survey.

The selection of participants considered their connection to the institutional context of the investigated object. We sought to obtain specific perceptions regarding the use of Generative Artificial Intelligence, focusing on efficiency, procedural speed, and the quality of activities developed with the support of technology in the MPF.

The questionnaires applied allowed the collection of quantitative data on participant perceptions. The questions addressed the efficiency, procedural speed, and quality of activities developed with the support of Generative Artificial Intelligence, according to the instrument detailed in the appendix.

The research was conducted in accordance with the ethical principles established by the USP/Esalq regulations. All participants were previously informed about the objectives of the study, the voluntary nature of their participation, the confidentiality of their responses, and the exclusive use of the data for academic purposes.

In observance of the General Law on Personal Data Protection (Law No. 13.709/2018), names, positions, or any information that could identify the respondents or their work units were not disclosed. The protection of personal data and the anonymity of participants were ensured throughout the collection and analysis process.

The research was conducted in sequential steps. Initially, the object of study was delimited, identifying Fiscal Representations for Criminal Purposes (RFFP) as the central instrument of tax criminal prosecution and the phases of its processing in the Federal Public Prosecutor’s Office.

Next, a normative and doctrinal survey was carried out, covering the applicable tax and criminal legislation, institutional regulations, technical manuals, guidelines from the Brazilian Federal Revenue Service, and academic studies related to RFFP and the use of Artificial Intelligence in the public sector.

Subsequently, Generative Artificial Intelligence tools were applied to support the reading, organization, cross-referencing, and synthesis of information contained in the RFFP. This procedure aimed at identifying indications of materiality, authorship, and criminal typicity, as part of the analysis of the object of study.

Subsequently, institutional perceptions were collected by applying structured questionnaires to members and interns of the Federal Public Prosecutor’s Office. The objective was to capture perceptions about the efficiency, speed, and quality of institutional performance with the support of Artificial Intelligence.

Finally, a comparative analysis of the obtained results was performed. Traditional RFFP analysis procedures were confronted with those supported by Generative Artificial Intelligence tools, evaluating the impacts on work rationalization and decision-making support.

3. Results and Discussion

The research conducted at the Federal Public Prosecutor’s Office (MPF) focused on analyzing the application of Generative Artificial Intelligence in the context of Fiscal Representations for Criminal Purposes (RFFP), seeking to understand its impact on procedural efficiency and speed. Initial findings revealed that the technology is perceived as a relevant support tool for the performance of institutional activities, especially regarding the management of large volumes of data and the drafting of legal documents, albeit with the caveat of indispensable human supervision.

The study sample consisted of seven participants, four of whom were members of the Federal Public Prosecutor’s Office, corresponding to 57.1% of the total, and three interns, representing 42.9%. This composition allowed for an analysis that favored the institutional perspective, given the predominance of members, while also incorporating the complementary view of the interns, who are directly involved in supporting procedural activities. It is important to note that the research did not include servers from the MPF, which delimits the scope of the perceptions collected.

The respondents’ perceptions on the use of Generative Artificial Intelligence were systematized into analytical categories, evidencing a significant consensus in some aspects. The entirety of participants, i.e., 100%, agreed that Generative Artificial Intelligence aids in the organization and analysis of data, and that there is an unquestionable need for human review of the content produced. These two points represent the central pillars of the perception of the tool within the MPF.

The assistance in data organization and analysis, with 100% agreement among participants, is a particularly relevant finding. It demonstrates the unanimous recognition of Generative Artificial Intelligence’s capability in handling the large documentary volume inherent to Fiscal Representations for Criminal Purposes, facilitating the synthesis of complex information and content triage. This perception aligns with literature highlighting AI’s role in Knowledge Management, supporting the creation, retrieval, and application of information in complex organizational environments, as pointed out by the National Council of the Public Prosecutor’s Office (2023) in its initiatives.

The relevance of Generative Artificial Intelligence for institutional activities and the optimization of time and productivity were perceived by six of the seven respondents, totaling 85.7%. This expressive percentage indicates broad acceptance of the technology as a valuable tool to support legal and administrative work. The ability to reduce the time dedicated to repetitive and operational tasks, freeing up members and teams for more analytical and strategic activities, is one of the most highlighted benefits, converging with studies that point to AI as a means of increasing organizational efficiency (Kaplan and Haenlein, 2019).

However, it is crucial to emphasize that the present study was based on the participants’ perceptions and did not perform an objective measurement of analysis time, productivity, or correction rates. Therefore, findings related to time and productivity optimization should be interpreted as respondents’ subjective view of the tool’s potential, rather than as empirical data of measured performance.

The standardization and quality of legal documents were also positively impacted by Generative Artificial Intelligence, according to 71.4% of respondents. This result suggests that the tool is seen as an instrument capable of conferring greater uniformity to institutional analyses and products, such as reports and drafts. Although there was no unanimity, the percentage is significant and indicates a favorable perception regarding the use of technology in the preparation of technical texts, in line with international experiences that use AI to assist in the production of standardized responses (Organisation for Economic Co-operation and Development, 2023).

One of the most crucial findings with 100% agreement was the need for a comprehensive human review of content generated by Generative Artificial Intelligence. This central research finding demonstrates that, despite being perceived as useful and promising, AI is not seen as a substitute for human action. There is an absolute consensus among participants that the content produced must be reviewed and validated, especially due to the requirement for legal accuracy, functional responsibility, and adherence to technical criteria.

This perception finds strong support in international and national ethical guidelines. UNESCO (2021) emphasizes that the use of artificial intelligence systems must be based on principles such as transparency, accountability, and human oversight, in order to mitigate risks related to errors and biases. The Organisation for Economic Co-operation and Development (2019) and the European Commission (2019) reinforce the need for human control to ensure reliable and safe decisions. In Brazil, the General Data Protection Law (Brasil, 2018) and the National Council of the Public Prosecutor’s Office (2023) also establish governance and transparency guidelines for the use of AI, especially given the risks of decision automation.

Finally, the category “challenges (training, access, and reliability)”, pointed out by 71.4% of respondents, reveals that the incorporation of Generative Artificial Intelligence in the public sector is not limited to the technical availability of the tool. Significant obstacles remain related to user training, standardization of access to tools, and the reliability of the results produced. Building a public workforce prepared for artificial intelligence requires continuous investment in training and the establishment of governance structures and clear guidelines for compliance and appropriate use (Organisation for Economic Co-operation and Development, 2021; Meijer and Wesseling, 2019).

In summary, the use of Generative Artificial Intelligence in the Federal Public Prosecutor’s Office demonstrated a significant contribution to institutional efficiency and speed in the analysis of Fiscal Representations for Criminal Purposes, mainly by optimizing data organization and analysis and standardizing the drafting of documents. However, this contribution is intrinsically conditioned by rigorous human supervision, continuous user training, and the implementation of clear governance guidelines, ensuring that technology acts as strategic support and not as a substitute for legal expertise.

4. Conclusion

This study analyzed the use of Generative Artificial Intelligence in the Federal Public Prosecutor’s Office, with an emphasis on its application in the analysis of Fiscal Representations for Criminal Purposes (RFFP) as a tool to support institutional efficiency and procedural speed. It was verified that the technology is perceived as a relevant tool for the performance of activities, especially in assisting the organization and analysis of large volumes of data, as well as in the standardization and quality of legal documents. The findings indicated that Generative Artificial Intelligence promoted gains in procedural analysis speed and assisted in the systematization of complex information, directing focus towards analytical activities. However, the unanimous need for human supervision to validate the produced information was observed, given the requirement for legal accuracy and the functional responsibility inherent in the ministerial role. The study’s main contribution lies in demonstrating that the incorporation of Generative Artificial Intelligence represents a significant advance in optimizing workflows and improving the technical quality of institutional manifestations.

Despite the identified benefits, the study presented limitations stemming from the reduced size and specific composition of the sample, which did not include servers, and from the concentration of the research in a single institution, which restricts the generalization of the results. Furthermore, the findings reflect participants’ perceptions, without objective measurement of analysis time or productivity. Challenges persist regarding user training, standardized access to tools, and result reliability, indicating that the adoption of these technologies demands an integrated approach involving governance, professional qualification, and alignment with ethical, legal, and administrative principles. For future studies, it is recommended to conduct objective comparisons between the traditional processing of RFFP and that supported by Generative Artificial Intelligence, using performance indicators, and to propose governance guidelines for the use of AI that ensure legal quality and human supervision as central elements.

Bibliographic References

BRASIL. Lei nº 13.709, de 14 de agosto de 2018. Lei Geral de Proteção de Dados Pessoais (LGPD). Diário Oficial da União: Brasília, DF, 15 ago. 2018.

BRASIL. Lei nº 14.129, de 29 de março de 2021. Dispõe sobre princípios, regras e instrumentos para o Governo Digital e para o aumento da eficiência pública. Diário Oficial da União: seção 1, Brasília, DF, 30 mar. 2021.

Chizzotti, 2003 [Referência completa não encontrada no documento original]

CONSELHO NACIONAL DO MINISTÉRIO PÚBLICO (CNMP). Relatório de iniciativas de Inteligência Artificial no Ministério Público brasileiro. Brasília: CNMP, 2023.

COSTA, R. F. Inteligência artificial e eficiência no Poder Judiciário: perspectivas de transformação digital. Revista de Direito e Tecnologia, v. 8, n. 2, p. 15-32, 2020.

EUROPEAN COMMISSION. Ethics Guidelines for Trustworthy Al. Brussels: European Commission, 2019.

GIL, A. C. Métodos e técnicas de pesquisa social. 6. ed. São Paulo: Atlas, 2008.

KAPLAN, Andreas M.; HAENLEIN, Michael. Siri, Siri, in my hand: Who’s the fairest in the land? On the interpretations, illustrations, and implications of artificial intelligence. Business Horizons, v. 62, n. 1, p. 15-25, 2019.

MARTINS, G. A.; THEÓPHILO, C. R. Metodologia da investigação científica para ciências sociais aplicadas. São Paulo: Atlas, 2009.

MEIJER, Albert; WESSELING, Jeroen. Institutional capacity for Al in government. Government Information Quarterly, 2019.

MINISTÉRIO PÚBLICO FEDERAL (MPF). Estratégia de Inovação e Transformação Digital. Brasília: MPF, 2023.

ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT (OECD). Building an Al-ready public workforce. Paris: OECD, 2021.

ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT (OECD). Governing with Artificial Intelligence: Are governments ready? Paris: OECD, 2023.

ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT (OECD). OECD Principles on Artificial Intelligence. Paris: OECD, 2019.

STRYKER, J.; KAVLAKOGLU, E. What is artificial intelligence (AI)? IBM, 2024. Disponível em: https://www.ibm.com/topics/artificial-intelligence. Acesso em: 26 out. 2025.

UNESCO. Recomendação sobre a Ética da Inteligência Artificial. Paris: UNESCO, 2021.

Article originating from the Course Conclusion Work of the Specialization in Tax Management of the MBA USP/Esalq

To learn more about the course, click here and access the MBX Academy platform