evidoria

← Back to browse

Good practice Imported

ChatTCU — Brazil's Federal Audit Court Logs 1.3 Million Internal AI Chats, With Mandatory Human Review

Brazil · Brasília · See the Brazil profile · See the Brasília profile

Top 25% 73/100 · Ask Evidence Copilot about this practice

Brazil's federal audit court built ChatTCU, an internal generative-AI assistant on Azure OpenAI with retrieval over its own case law. By June 2026 it logged 1.3 million chats among 3,860 staff, with mandatory human review of every output before use.

ChatTCU — Brazil's Federal Audit Court Logs 1.3 Million Internal AI Chats, With Mandatory Human Review

Details

Promoter
Tribunal de Contas da União (TCU)
Period
February 2023 – June 2026
Keywords
public audit, internal government operations, generative AI

Description

The Tribunal de Contas da União (TCU), Brazil's federal audit court, launched ChatTCU in February–March 2023 as an internal chatbot built on OpenAI's GPT-3.5 model. A second version followed in June 2023, and a third in October 2023 added retrieval-augmented generation (RAG) over TCU's own internal systems — case law, prior rulings and administrative records — rather than relying only on the underlying model's general knowledge.
By June 2026, according to a Microsoft customer case study, ChatTCU had processed 1.3 million chats and 4.9 million messages among 3,860 distinct users, with typical daily volumes of 10,000–30,000 messages. Staff use it to search institutional knowledge in natural language, draft preparatory text for opinions and rulings (such as Voto Recon and Acórdão de Relação drafts), and support audit and financial analysis — TCU officials Rainério Rodrigues Leite (IT Secretary) and Wagner Miranda Costa (Deputy Technology Secretary) describe the goal as giving staff access, via generative AI, to institutional data that public AI models cannot see.
Governance safeguards documented by TCU and Microsoft include mandatory human review of every AI-drafted output before it is finalised, access restricted to TCU personnel, encryption at rest and in transit, and a dedicated network link with no public internet exposure. TCU's own domestic AI-maturity survey and OECD commentary have separately flagged that most federal Brazilian institutions remain at early stages of AI adoption, making TCU's own advanced internal use an outlier rather than the norm.
An honest limitation: the public figures describe usage volume and technical safeguards, not an independent evaluation of output accuracy, error rates, or measured time savings per case — those effectiveness metrics have not been published.

Read the full analysis: https://portal.tcu.gov.br/imprensa/noticias/uso-de-inteligencia-artificial-aprimora-processos-internos-no-tribunal-de-contas-da-uniao

Implementation

Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful