nalysis of the Technical/Regulatory Process for the Adoption of LLM Tools in the Brazilian Public Prosecutor’s Office
Public Administration; Artificial Intelligence; Large Language Models; Public Prosecution Service; Algorithmic governance.
The digital transformation of Public Administration has incorporated artificial intelligence systems, especially Large Language Models (LLMs), into service provision, information analysis and decision support activities. In Brazil, the Public Prosecution Service is a strategic setting to observe the opportunities and risks associated with these technologies, particularly regarding data protection, transparency and algorithmic governance. This study aims to analyse how LLMs are being implemented and regulated within the Brazilian Public Prosecution Service, as well as which institutional capacities are being mobilised to ensure their responsible use. An exploratory, qualitative research design is adopted, based on a case study of the Public Prosecution Service and combining primary data, collected through a semi-structured questionnaire sent to prosecutorial units, with secondary data gathered from official portals and normative repositories. The data will be examined through content analysis, using categories related to digital bureaucracy, data analytics and public value. The study is expected to map operational and managerial uses of LLMs, identify technical, organisational and regulatory barriers, and detect gaps in internal guidelines and protocols. In doing so, it seeks to provide input for improving policies and regulations that guide the ethical and transparent use of LLMs, in line with constitutional principles, within the Public Prosecution Service and, more broadly, in Brazilian Public Administration.