He watched a hologram of himself promise victory to a crowd while the courthouse clock kept ticking. You could feel the mismatch: a man under court-ordered silence, speaking to thousands via code. The projection landed like a political provocation and the judges have noticed.
I’ve been tracking how AI reshapes rallies and rulebooks, and this one is an accelerated test case. You’ll recognize the players: Jair Bolsonaro, under house arrest after a 27-year sentence; his son Flávio, launching a presidential bid; and a giant AI-generated projection that said, in plain Portuguese, “I’m silenced but hopeful.” The video lives on YouTube, and the lawyer’s denial appeared on X.
The stunt felt theatrical: a puppet stitched from code on a stage built of rules.
At the rally, a projected Bolsonaro spoke as if he were free
The projection opened with a confession: it labeled itself as artificial and then spoke in unmistakable Bolsonaro cadence. That admission is central to the defense offered by Flávio’s campaign—“What you see here is not me. This is a simulation of my image and my voice using artificial intelligence.”
But declarations don’t erase consequences. Bolsonaro is serving a 27-year sentence and has been moved to house arrest because of health concerns. One explicit condition of his relocation was a ban on public communication, and his lawyer, Paulo Cunha Bueno, says the former president neither authorized nor could have authorized the AI clip because he’s restricted from visitors. If the father’s lack of consent is accepted as fact, the son’s campaign could be flirting with criminal exposure under Brazilian electoral law.
Outside the courtroom, the election authority already has rules aimed at this
The Tribunal Superior Eleitoral (TSE) in Brazil has clear language banning the political use of AI deepfakes without disclosure. That makes this more than theater: it’s a legal test.
The TSE now has to decide whether a labelled deepfake—one that begins by saying it is artificial—satisfies the law or if the absence of consent from the person portrayed changes the calculus. Legal scholars warn the court could disqualify Flávio’s candidacy if it concludes the stunt violated electoral rules. The Financial Times has noted the TSE’s options include sanctioning the campaign or worse: removing a name from the ballot.
Can AI deepfakes be illegal in Brazil?
Yes—Brazil’s electoral law already targets misleading or manipulative AI content used for political ends. The law doesn’t only care whether something is labeled; it asks about consent, intent, and the likely effect on voters. The TSE must weigh disclosure against authorization and the specific context of a jailed politician who cannot lawfully communicate.
Could Flávio Bolsonaro be disqualified for using a deepfake?
It’s possible. Courts consider the spirit of electoral fairness. If the TSE finds the campaign crossed the line—using an unauthorized replica of a silenced public figure—then disqualification is on the table. That outcome would reshape how campaigns use synthetic media across Brazil.
In feeds and comment threads, platforms became the battleground
The video was shared widely on YouTube and amplified on X faster than judges could react. Platform labels and takedown policies are now acting as stopgaps between viral spectacles and official remedies.
The European Union’s AI Act already requires that generated or manipulated media be clearly disclosed—an anchor the TSE’s rule echoes. The United States lacks a uniform federal disclosure rule, leaving platforms like YouTube and X to set their own lines. Meanwhile, the same voice-cloning and generative-video tools—commercial services such as Synthesia and open frameworks like DeepFaceLab—make it cheap and fast to produce content that looks and sounds real.
The crowd reaction made the stakes plain: voters felt something was returned to them
People in the square cheered as if they’d heard Bolsonaro live. For supporters, that sensation is potent; for opponents, it’s a demonstration of how synthetic media can short-circuit democratic norms. Critics warn politicians may use such content to curate images while later discrediting real evidence as “just AI.”
This is a political cat-and-mouse: a campaign gambit that can look like empathy or manipulation depending on who’s watching. The tactic functions like a Trojan horse with a megaphone—small inside, enormous once it opens.
I want you to imagine the precedent the TSE sets here: will courts treat labeled, unauthorized AI as protected speech or as an illicit campaign tool? The decision will ripple across platforms, influence lawmaking from Brasília to Brussels, and define whether synthetic likenesses become routine in election theaters—so what would you do if your political truth could be cloned and staged for millions?