Regulating the Regulation: How the TCU Exercised Its Independence to Address Artificial Intelligence Challenges in the Brazilian Public Sector
Authors: Diego Farias; Klauss Nogueira; Pedro Coutinho Filho; Thacio Scandaroli, auditors of the Specialized Information Technology Audit Unit of TCU (Brazilian Federal Court of Accounts).
1. Introduction: The Crossroads of Innovation and Independence
The rapid expansion of artificial intelligence (AI) in the public sector has placed Supreme Audit Institutions (SAIs) at a new frontier of oversight. Amid promises of increased efficiency and significant risks to ethics, transparency, and accountability, the role of SAIs demands technical discernment and institutional autonomy.
The Brazilian Federal Court of Accounts (TCU) has assumed a practical leadership role in this context. Beyond implementing AI in its own operations, the TCU has actively influenced the public sector ecosystem by sharing the source code of its tools with other institutions. One example is ChatTCU—a generative AI-based support tool—whose code was made publicly available to allow other agencies to develop customized applications.
TCU’s leadership also extends to the regulatory field. In 2024, the Court submitted the executive summary “Risks of Regulating Artificial Intelligence in Brazil” to the National Congress, warning of potential negative effects of poorly calibrated regulation and advocating for balanced and technically sound approaches. The publication was widely disseminated and helped solidify TCU’s role as an independent and credible voice in the governance of emerging technologies.
This leadership also encompasses other emerging technologies. TCU published an executive summary on blockchain, evaluating both risks and potential in the public sector. These publications, along with concrete actions for sharing knowledge and technology, illustrate how TCU exercises its independence constructively, offering technical solutions even in uncertain regulatory environments.
This article presents TCU’s experience as an example of how independence is not only resistance to external pressures, but also the ability to act with autonomy, technical consistency, and a sense of public responsibility in the face of technological challenges.
2. TCU as a Technical Leader: Use and Dissemination of AI in the Public Sector
TCU has emerged as a national and international reference in the use of AI to enhance audit and oversight activities. Since the launch of ChatTCU in March 2023, the tool has been widely adopted internally, assisting auditors and staff in improving processes and efficiency. By May 2025, ChatTCU had over 2,700 active users, representing around 90% of the institution’s workforce.
Beyond internal use, TCU has taken a collaborative approach by sharing ChatTCU’s source code with other public institutions. As of mid-2025, 136 institutions—including public agencies, regulatory bodies, professional councils, universities, and audit courts—had signed agreements to adopt the technology. The initiative also led to a formal technology transfer agreement with the SAI of Honduras and the Office of the Comptroller General of Chile.
This reflects TCU’s commitment to innovation and its role as a catalyst for digital transformation in the public sector. By enabling widespread access to ChatTCU’s foundations, the Court reinforces its technical leadership and collaborative ethos.
TCU’s international recognition further supports its position. In April 2024, the Organisation for Economic Co-operation and Development (OECD) recognized TCU as the only government entity to reach an advanced stage in the use of generative AI, citing ChatTCU as a model of innovation and efficiency.
In 2025, Rainério Leite, TCU’s Secretary for IT Oversight and Digital Evolution, was included in the “Government AI 100” list as one of the world’s 100 most influential individuals in AI applied to the public sector. This recognition underscores TCU’s leadership in balancing innovation, ethics, and effective public sector oversight. Its approach is rooted in principles of algorithmic ethics, transparency, and accountability, which guide both internal development and the external evaluation of AI tools.
3. Blockchain and the expansion of TCU’s regulatory role
TCU’s innovative and independent posture is also evident in its engagement with other emerging technologies, such as blockchain. In 2020, the Court released the executive summary “Mapping Blockchain Technology,” which provides a framework for guiding public institutions in developing blockchain projects, based on national and international experience.
The publication emphasized that blockchain adoption should follow a feasibility assessment, supported by cost-benefit analysis, problem alignment, and strong governance structures. It also warned of the risks of ineffective public investment if blockchain were adopted solely for its novelty, without solid technical justification.
One key risk highlighted was the lack of a consolidated legal framework, which can lead to legal uncertainty, diverging interpretations, and misallocation of public funds. The document recommends conditioning blockchain use on demonstrable public value and careful planning of interoperability, security, and sustainability requirements.
TCU also stressed that blockchain is not a universal solution. Its applicability depends on specific conditions, such as the need for immutable records and decentralized trust.
Notably, the report anticipated the potential role of the federal government as a provider of public infrastructure for blockchain solutions. This vision materialized in April 2022 with the creation of the Brazilian Blockchain Network (RBB), coordinated by the federal data processing service (Serpro) and the national development bank (BNDES), with participation from various government bodies.
This pioneering initiative reinforces TCU’s role as a promoter of sound regulatory practices, even in uncertain and evolving technological landscapes, and exemplifies how institutional independence can be exercised through the production of qualified public knowledge.
4. The publication “Risks of Regulating AI in Brazil”
In April 2024, TCU presented to the National Congress the executive summary “Risks of Regulating Artificial Intelligence in Brazil,” a technical document prepared to support the legislative debate on the topic by presenting and discussing key regulatory risks and their potential impacts on Brazil’s national AI strategy.
The publication identified nine main risks, which can be grouped into four categories:
- Restriction of innovation and digital transformation – includes disincentives to innovation, delays in implementation, and limits to the digital transformation capacity of the state.
- Market concentration and exclusion of small actors – includes barriers to startups, regulatory advantages for monopolies or oligopolies, and challenges in retaining AI talent.
- Poorly calibrated regulation and legal uncertainty – includes outdated or ineffective regulatory models, generic legal definitions that affect unrelated areas, and rules incompatible with AI realities (e.g., copyright law).
- Institutional and international misalignment – includes conflict among regulatory bodies, lack of technical capacity among regulators, and misalignment with international best practices.
Using clear and accessible language, and accompanied by an explanatory infographic, the TCU publication helped raise awareness of the risks posed by poorly designed regulation. It did not propose new rules, but rather acted as a technical alert and institutional contribution to public governance. Its dissemination in a politically sensitive context reaffirmed TCU’s role as a neutral yet proactive institution.
5. Dilemmas faced: between innovation and external pressures
TCU’s AI initiatives brought visibility and praise but also significant institutional challenges. The publication on AI regulatory risks was welcomed by technical sectors but sparked divergent reactions from lawmakers and executive officials advocating stricter or more centralized regulation.
In this context, the Court had to balance a proactive stance with the need to maintain technical neutrality. By advocating a prudent, evidence-based regulatory approach, TCU mitigated reputational risks and avoided the perception of political bias.
Internally, the deployment of tools like ChatTCU also required caution to avoid perceptions of self-promotion or undue reliance on technology over traditional methods. The Court addressed these risks by releasing source code, hosting public debates, and opening its strategies to expert and international critique.
These dilemmas illustrate that independence is not merely a legal or formal concept—it is a continuous exercise in institutional judgment, especially when innovation challenges established expectations.
6. Paths forward: Independence as a regulatory asset
TCU’s experience in the field of artificial intelligence points to valuable lessons for other SAIs facing similar challenges amid regulatory and technological change.
First, action in emerging regulatory environments requires a combination of technical knowledge, strategic timing, and institutional prudence. TCU’s leadership in producing guidance—such as the executive summary on AI regulation—shows that it is possible to influence public policy while preserving neutrality and institutional credibility.
Another replicable practice is algorithmic transparency. By releasing the ChatTCU source code and supporting its adaptation by other institutions, TCU fosters a culture of cooperation while avoiding technological monopolies. This shows that independence means having the autonomy to lead responsibly, not acting in isolation.
Finally, active listening to civil society and dialogue with experts, universities, and international organizations strengthen the legitimacy of oversight actions. These strategies affirm the value of independence as a regulatory asset—a necessary condition for SAIs to not only oversee but also contribute solutions to the most complex public challenges.
7. Conclusion
TCU’s recent experience demonstrates that institutional independence is not only about resisting external interference, but about leading technically, contributing to public policy, and acting with consistency between discourse and practice.
By combining the responsible use of AI with a critical and collaborative stance toward emerging regulation, the Court reaffirmed its functional autonomy and its commitment to the public interest. Its publications and technical analyses on AI regulation and blockchain adoption in public administration are concrete expressions of an independence that looks outward, benefiting the broader public ecosystem.
TCU’s case illustrates that leadership in sensitive areas, when exercised with prudence, transparency, and technical grounding, can enhance the legitimacy and impact of Supreme Audit Institutions. In a rapidly changing world, maintaining independence means more than resisting—it means being prepared to act with judgment and courage when inaction would be too costly.
References
- TCU supports institutions in implementing generative artificial intelligence tools. Available at: https://portal.tcu.gov.br/imprensa/noticias/instituicoes-parceiras-recebem-suporte-tecnico-para-implementacao-do-chattcu
- TCU and the Office of the Comptroller General of Chile sign an agreement for cooperation in technology and AI: https://portal.tcu.gov.br/imprensa/noticias/tcu-e-controladoria-geral-do-chile-firmam-acordo-para-cooperacao-em-tecnologia-e-ia
- OECD recognition – Generative AI for anti-corruption and integrity in government. Available at: https://www.oecd.org/en/publications/generative-ai-for-anti-corruption-and-integrity-in-government_657a185a-en.html
- Government AI 100 Award – Apolitical. Available at: https://apolitical.co/list/en/government-ai-100-2025
- Executive Summary – Mapping Blockchain Technology. Available at: https://portal.tcu.gov.br/data/files/59/02/40/6E/C4854710A7AE4547E18818A8/Blockchain_sumario_executivo.pdf
- Risks of Regulating Artificial Intelligence in Brazil – Executive Summary. Available at: https://portal.tcu.gov.br/publicacoes-institucionais/sumarios-executivos/riscos-da-regulacao-da-inteligencia-artificial-no-brasilBrazilian Blockchain Network. Institutional Presentation. Available at: https://www.gov.br/rbb
