Senate and Receita Federal Tap Embrapii to Train Civil Servants in AI
Two of Brazil’s most data-heavy federal institutions are sending staff back to class. The Federal Senate and the Receita Federal have signed agreements with the Embrapii Competence Centre in Immersive Technologies, known as Akcit, at the Federal University of Goiás (UFG), to train civil servants in artificial intelligence and immersive technologies. Embrapii announced the partnerships on October 2.
The plan includes graduate and specialisation programmes in generative AI, natural language processing, software engineering and agentic AI, with 500 places offered by the centre. Embrapii stresses that the goal goes beyond using off-the-shelf tools: staff are expected to learn how to build and adapt models and applications for their own institutions.
The intended uses are concrete. In the Senate, the skills could support document analysis, the organisation of legislative information, database research and tools for technical staff, as well as easier public access to bills and debates. At the Receita, training could improve the analysis of tax and customs data, flag inconsistencies, automate repetitive tasks and strengthen fraud prevention. Embrapii president Alvaro Prata said training public servants in advanced technologies gives the state more capacity to identify problems and develop its own solutions.
What it means in practice
The hardware behind the courses is substantial for a university lab. Akcit’s Advanced Laboratory of Immersive Technologies includes an NVIDIA DGX B200 supercomputer with eight GPUs, part of roughly R$ 2.5 million in investment, and the centre is linked to UFG’s Centre of Excellence in Artificial Intelligence (Ceia). Companies including Globo, Itaú Unibanco, WEG and Vivo already take part in its technology association.
For federal managers, the move fits a wider push to build state AI capacity instead of renting it. AI4Brazil has tracked how EAF is studying a modular public AI data centre for federal use and how the ANPD is building an internal AI centre. Trained staff are the missing piece that makes that kind of infrastructure useful.
Embrapii itself notes the caveats: using AI on legislative and tax data demands security, data protection, transparency and human oversight. Tax enforcement is an especially sensitive area, where automated flags must stay explainable and open to challenge by taxpayers.
What to watch: when the first cohorts start, whether other ministries adopt the Akcit model, and whether the Senate and the Receita publish the AI tools their staff build. Five hundred places will not transform the federal workforce on their own, but they are a concrete step toward a public sector that understands the systems it buys.