impactRIO

impactRIO is a Rio de Janeiro-based AI-safety study and research hub, self-described as "the first AI safety group in Brazil." Founded in July 2023 by João Lucas Duim, it runs structured fellowships based on BlueDot Impact's AI Safety Fundamentals curriculum — covering alignment, advanced alignment, and governance tracks — and hosts monthly EA meetups connecting Rio de Janeiro's effective-altruism community. AISafety.com's global community directory lists impactRIO as Active and places it at Fundação Getulio Vargas (FGV), one of Latin America's leading research institutions. As of the group's June 2024 one-year update, 26 people had completed at least one fellowship and 9 had changed their career plans toward AI safety work.

Founding and structure

impactRIO was founded in July 2023 by João Lucas Duim, who serves as director. The group began as an unofficial university club, initially supported by UGAP (University Group Accelerator Program), the Condor Initiative, EA Brazil, and OSP AIS (a predecessor to the Fieldbuilder Support Program). The university withdrew its institutional support approximately one year after founding; subsequent to that transition, AISafety.com places impactRIO at Fundação Getulio Vargas. The only comparable simultaneous founding effort in Brazil, ALAI USP at the University of São Paulo, was noted as inactive by mid-2024, leaving impactRIO as the primary active AI-safety community group in Brazil.

Programs

impactRIO's core program format is the structured cohort fellowship built around BlueDot Impact's curriculum:

Alignment fellowship (2023). A ten-session cohort based on the AI Safety Fundamentals curriculum — the standard entry-point track for the EA-adjacent AI-safety field-building pipeline.

Advanced Alignment fellowship (2024). A ten-session cohort based on BlueDot's Alignment 201 curriculum, targeting participants who had already completed the Fundamentals track.

Governance fellowship (2024). A parallel ten-session cohort based on BlueDot's AI Governance curriculum, running concurrently with the Advanced Alignment track. The two 2024 cohorts drew satisfaction ratings of 9.7/10 (Governance) and 9.3/10 (Advanced Alignment).

Beyond fellowships, impactRIO runs monthly EA meetups in Rio de Janeiro — a sustained effort to build connections among existing EAs in the city and introduce new participants to EA ideas. Guest speakers from the full-time AI-safety research community gave fellowship participants direct access to practitioners in a city where Brazil's thin AI-safety researcher base (~five full-time researchers nationally as of mid-2024) is concentrated. The LinkedIn page notes participation in ML4Good Brasil 2025, a machine-learning-for-social-good bootcamp series.

Outcomes

The June 2024 update reported 26 fellowship completions across the three cohorts, with organisers and members collectively logging at least 27 attendances at international EA and AI-safety events. Nine people changed their career plans to focus on AI-safety risk mitigation; two more accelerated pre-existing safety-related career tracks. Of those who applied to Condor Camp (an AI-safety residential programme), five of eleven were accepted; three of nine were accepted to EA Global; three of eight to ML4Good. The average social comfort rating across programs was 9.6/10.

Regional connections

João Lucas Duim is one of eleven co-founders of the Latin American Network for AI Safety (LANAIS), which launched with a kick-off event on 28 June 2025 and aims to connect researchers, students, and organisers across Latin America through a public directory, career resources, and virtual and in-person networking events. The co-founders also include Eitan Sprejer, founding director of BAISH (Buenos Aires AI Safety Hub), establishing a direct personal link between the two most documented Latin American AI-safety hubs in the corpus. LANAIS is a pan-regional network, not a single hub, and is organisationally distinct from impactRIO.

Place in the movement

impactRIO performs the insider-community pattern characteristic of the EA-adjacent AI-safety field-building layer: structured cohort fellowships, small peer groups, and pipeline development toward research and governance careers, rather than public-facing campaign organising. Its 2023 founding and early-support structure — UGAP, Condor Initiative, EA Brazil — are standard markers of this layer, placing it alongside BAISH in Buenos Aires, AI Safety South Africa in Cape Town, and the other Global South community groups the corpus already holds.

What distinguishes impactRIO within the Latin American layer is its Rio de Janeiro context and its specific role as Brazil's first AI-safety community group. Brazil has its own national AI governance arc — the ANPD data-protection authority, draft AI regulation legislation, and civil-society AI-safety institute proposals by CEBRI and ITS Rio — yet the grassroots practitioner community in Rio de Janeiro that could feed into that governance infrastructure remained thin through impactRIO's first two years. The group's June 2024 priorities — a formal research group, an alignment/governance hackathon, and integration of AI-safety content into undergraduate curricula — describe the same development trajectory BAISH has moved further along in Buenos Aires, suggesting impactRIO's trajectory is early-stage relative to its regional peers.