An Artificial Intelligence Hub in Brazil’s Federal Public Accounts Tribunal: Advances and Challenges in Oversight and Control
Brazil's Federal Public Accounts Tribunal (TCU – Tribunal de Contas da União) has found a powerful ally in artificial intelligence (AI) for optimizing and reducing time spent in preparing texts, translations, and analyses for its oversight activities.
The creation of its AI hub, NIA (Núcleo de Inteligência Artificial) marks a new phase for TCU. NIA's mission is to develop and implement AI solutions that focus on improvements to public management and citizen services.
"ChatTCU", the AI tool developed by the TCU, is secure because it is used in a restricted environment. In addition, interactions with ChatTCU within the environment are protected, ensuring that information provided by users will be kept confidential.
The TCU stands to make significant gains through the use of AI, such as the automation of complex processes and detection of patterns in large volumes of data. The initiative is aligned with one of the TCU's institutional guidelines for oversight activities, which prioritizes intensive use of technology.
Allied to AI, machine learning optimizes work processes and resource allocation, and the TCU plans to increases its investments in this technology over the coming years.
According to Directive-Setid no. 2/2023, NIA's responsibilities include:
Recommending, researching, evaluating, defining, deciding on, and supplying new AI-based solutions;
Providing support to TCU users in the use of AI services;
Conducting and supporting initiatives for communication and awareness, adoption of best practices and "lessons learned" by project teams in the use of AI-powered solutions;
Collaborating with other TCU administrative units on data mining, machine learning, natural language processing, and other data analysis methods; and
Creating rules and guidelines for use of AI services.
The creation of the TCU's AI hub is part of the wave artificial intelligence solutions in government – in fact, the TCU conducted a survey (Acórdão 1139/2022 – Plenário – Portuguese only) to assess the use of AI by federal government offices and agencies.
Beyond its own activities, the TCU has been attentive to the increasing use of AI by government. In September 2021, TCU identified at least 75 federal organizations that actively employed AI-based solutions. No fewer than 44 AI tools were used by government entities: 36 for internal use and only 8 dedicated to serving citizen needs directly. Various AI projects were identified in the Judiciary, with 29 under development and 27 operational in 2021.
The use of AI tools in oversight activities is not new to the TCU. The tribunal has developed several AI-based solutions in recent years, focusing on planning, conducting audits, and generating and managing information and evidence in proceedings. Examples include robots like Alice, Mônica, Adele, and Ágata, along with tools like Detecta and Sofia.
However, the TCU's increasing use of AI also presents challenges and potential risks. The reliability and security of AI-based systems must be continually monitored, since algorithmic biases or failures can lead to incorrect or unfair decisions. Privacy of citizen data and the TCU's operational transparency must also be considered, given that AI technologies involve collecting and processing large amounts of sensitive information. As the TCU moves forward with AI implementation, therefore, rigorous monitoring is essential to mitigate these risks and ensure that the technology's benefits are harnessed responsibly.
