PT
The start of free radio and TV campaign airtime opens the season of fighting disinformation and searching for paths through investment in artificial intelligence
RADAR PUPA — Aug 28, 2026
By José Brito, journalist, founder of Pupa Educação Digital and teacher at the AI and Digital Education Lab at Colégio MOPI.
“Social media and content created by artificial intelligence will generate some of the big stories of this election. The presence of candidates who matter in the polls, yet have no TV airtime, only reinforces that.”
Carlos Affonso Souza, director of ITS Rio, in his column for UOL..
Image generated by AI in Nano Banana. Pixel art: a tower of servers with an “AI” sign over the map of Brazil, surrounded by robots, with icons of a book, a mobile phone, a chip, a storm and a ballot box protected by a shield between two robot faces, one of them glitched.
Do you know what a petaflop is? I had never heard this odd little word, which sounds more like the name of a stomach-ache remedy than anything else. Well then. FLOP is the English acronym for floating-point operation. Not much help, right? OK. A petaflop is something like a quadrillion calculations per second, usually associated with climate simulations, molecular studies or even astrophysics research.
Brazil’s most powerful computer, affectionately named Santos Dumont, sits at the National Laboratory for Scientific Computing in Petrópolis, my hometown, in the Quitandinha neighbourhood. I grew up going with my father to volleyball matches at the Banco do Brasil Athletic Association, which stood on the same plot of land where our supercomputer now sits, capable of unimaginable feats with data and probabilistic models.
My childhood memories from the 1980s connected with the federal government’s announcement made in Rio Grande do Norte. The Technological Sovereignty package foresees around R$2.5 billion in investment in AI infrastructure, chip production and, of course, more resources for data centres. That is why this week I brought in the line from Carlos Affonso Souza — “Caff” — UOL columnist and a colleague of mine at ITS Rio, along with the warning about speeches and projects involving cutting-edge technology.
If today we are already struck by the alerts that arrive from Civil Defence on our phones — signalling far more than whether to take an umbrella when leaving the house, or whether it will be sunny at the weekend — imagine what lies ahead once we expand data processing and gain more precise access to information such as strong winds, floods, landslides, or even the gigantic waves of mud and ice like the ones seen in Nepal.
According to the LNCC’s own figures, the SDumont currently has an installed capacity on the order of 20 petaflop/s. To give you an idea, the new machine promises something close to 7,200 petaflops. If you didn’t quite follow, don’t worry. In theory, if all 8 billion people on the planet did one calculation per second, humankind would take 29 years to match what that computer processes.
While all of this is still being built — we are, after all, talking about a project for 2027–2030 — we need to pay attention to what is being said out there. It is election season, and AI-driven technology will dominate a significant part of the debates. Brazil’s Digital Child and Adolescent Statute (ECA Digital) has been in force since March 2026, and we are watching the ANPD, the National Data Protection Authority, oversee platforms amid the new court rulings that flood the news every week in Brazil and around the world. This time it was Meta that reached a billion-dollar settlement in the United States concerning screen use by teenagers.
MIT has published a report that opens up a rather interesting discussion: how do you segment rules for building policies on AI use? It applies to schools, it applies to the market, it applies to society. Before choosing tools, we need to understand the consequences of using them. The Superior Electoral Court (TSE) and Google have launched partnerships to fight deepfakes and AI-generated synthetic videos of candidates. Let’s follow where this debate goes, because today is only the first day of free campaign advertising on radio and TV.
📌 Sources: Data Center Dynamics | Observatório da Imprensa · Codesinfo | Fast Company Brasil | Jornal do Brasil · TSE and Google | MIT · AI and Education | ANPD · ECA Digital | Network World, 23/06/2026 | TOP500 | Movimento Econômico
#1 — Supercomputer and chip production in the sights of federal investment in AI infrastructure
A programme announced in Rio Grande do Norte foresees more than 2 billion reais in funding and partnerships with Chinese technology giants such as Huawei and iFlytek.
🔗Source: Data Center Dynamics | InfoMoney (link a inserir)
#2 — Meet five open-source tools for journalism routines powered by artificial intelligence
An innovation fund for journalism, backed by Google, supports free-of-charge solutions with a focus on mentoring teams and encouraging the use of the technology. Applications range from classifying news stories to reading official gazettes.
🔗Source: Observatório da Imprensa | Codesinfo
#3 — Meta’s US$18 billion settlement points to new rules for social media use by teenagers
Time limits, feed blocking and silenced notifications during school hours are among the measures for Instagram and Facebook users aged 13 to 17 in the US. The decision depends on court approval and does not automatically extend to Brazil.
🔗Source: Fast Company Brasil | coluna de Renata Cafardo · Estadão
#4 — TSE and Google launch campaign for candidates to detect deepfakes on YouTube
The “Likeness Detection” tool alerts registered candidates when a YouTube video reproduces their face, allowing them to request its removal. Institutions point to a greater risk for women candidates.
🔗Source: Jornal do Brasil
#5 — MIT recommends principles for drawing up AI-use policies for each school subject
The study reveals the importance of dialogue and transparency in shaping learning objectives with applied technology. Oral exams and project-based assessment gain greater relevance in building student autonomy.
🔗Source: MIT · AI and Education
READING TIP 📚
Book: Becos da memória (Alleys of Memory)
Author: Conceição Evaristo
Publisher: Pallas
Year: 2024
In times of screens everywhere, ask someone next to you: what are you reading? I heard that line the other day and decided to list some of my recent reads. Right now I’m with Conceição Evaristo, a Brazilian writer who holds chair 40 at the Academy of Letters of Minas Gerais and is one of the country’s great voices in the debate on the struggle for civil rights, ancestry, the fight against racism and the affirmation of diversity.
Becos da Memória runs through a little of all of that when read with attention and sensitivity to one of the concepts most widely spread by the author’s career: escrevivência, a word born from the union of “writing” (escrever) and “lived experience” (vivência), above all when tied to the collective struggle of Black women in a country marked by inequality.
Sovereignty also begins with access to literature. Access to the memory of what brought us here, and to a more plural and representative future for the Brazil we want.
Cover of Becos da Memória, by Conceição Evaristo, published by Pallas: against a dark purple background, an old photograph in orange duotone shows Black children and adults gathered and smiling, overlaid by a lattice of white diamonds. On the purple band at the bottom, the title in orange and the author’s name in white.
This is Radar Pupa, a weekly newsletter about learning with AI in education, digital culture and the future of work. If it makes sense for you, subscribe for free and share.