TSE Postpones AI Election-Punishment Decision as Global Deepfake Cases Pile Up

Brazil's top electoral court has postponed a meeting intended to create consensus on how to punish artificial-intelligence use in elections. The Tribunal Superior Eleitoral (TSE) was due to discuss whether the presidential campaign of Flávio Bolsonaro (PL) broke rules by using an AI-generated video of former president Jair Bolsonaro during the party convention that formalized his candidacy for the Palácio do Planalto. No new date has been set.

The delay leaves Brazil without a clear enforcement standard just as its official campaign period approaches. At the centre of the dispute is not whether the video exists, but what kind of AI-generated political content should draw sanctions, and how hard those sanctions should be.

Researchers say the same uncertainty is playing out globally. Sam Stockwell of the Centre for Emerging Technology and Security (CETaS) at the Alan Turing Institute says there is still no conclusive evidence that AI tools have decisively changed a national election result. Rather, the deeper harm is subtler: generated media can sow confusion about what is real, corrode trust in institutions and inflame political divisions.

Recent cases show the range of the problem. In Ireland, a fake RTÉ News broadcast showed presidential front-runner Catherine Connolly appearing to abandon her campaign three days before the vote; she nevertheless won with 63%. In Slovakia, a deepfake audio recording of liberal candidate Michal Šimečka circulated during the election silence period. In Pakistan, jailed former prime minister Imran Khan used AI-generated video from prison to declare that his allied independent candidates had won. And in Romania, deepfakes of presidential candidates were used not to tip the vote but to promote a fake investment scheme.

Advertisement

What Ireland, Germany, Slovakia and Romania Reveal About Deepfake Risk in Brazil

Why the TSE postponement creates immediate legal uncertainty

Without a new meeting date, Brazilian campaigns cannot know which AI uses will be punished, labelled or tolerated. The Flávio Bolsonaro case is the first high-profile test, but the court's eventual ruling could range from requiring disclosure labels to more serious campaign penalties. A delayed decision may encourage more aggressive synthetic media use during the official campaign, precisely when response time is shortest.

The fake-broadcaster pattern: RTÉ, CNN, France 24 and Deutsche Welle

Several operations relied on imitating trusted news brands rather than creating obviously fake outlets. Ireland's deepfake copied RTÉ News, while Ecuadorian videos faked the logos of CNN, France 24 and Deutsche Welle to place candidates in false corruption scandals. The tactic works by transferring trust in a real institution to a manipulated clip. Once the actual broadcaster denies the story, the false content has often already reached its audience.

Germany's false security alerts and the attribution problem

In Germany's February 2025 parliamentary campaign, AI narration was layered onto real police and MI6 footage to claim bomb threats, poisoned ballots and imminent attacks on voting centres. European institutions later traced the material to a Russian disinformation network. Stockwell says real institutions face a difficult balance: denying a video may inadvertently draw more attention to it, while staying silent risks allowing their credibility to be exploited.

No proven election flip, but measurable erosion of trust

Across Ireland, Ecuador and Slovakia, fact-checkers and scholars concluded the deepfakes did not demonstrably decide the result. In Slovakia, researchers argued pro-Russia susceptibility and political instability were stronger factors in Robert Fico's victory than the fake audio of Šimečka. But the Romanian case shows a different harm: deepfakes of candidates George Simion and Nicușor Dan promoted a 1,400 lei investment scheme with a promised 9,000 lei monthly return, turning voters into potential fraud victims at the height of political attention.

Advertisement

What Brazilian Campaigns, Voters and Platforms Should Do Before Voting Starts

The Brazilian case makes clear where immediate responsibility falls.

  • For campaign teams: prepare a verified-channel rapid response for a fake withdrawal or scandal clip, as Catherine Connolly's campaign did within hours of the RTÉ-style deepfake in Ireland. Have pre-drafted statements and a direct line to platform trust-and-safety contacts.
  • For voters: treat any video in which a candidate announces withdrawing from the race, endorses an investment, or declares victory before official results as high-risk. Check official TSE channels and the campaign's verified accounts rather than resharing first.
  • For platforms: the German case shows that real footage from police and intelligence agencies can be repurposed with AI narration. Content-moderation workflows should flag known institutional footage and remove or label manipulated election security alerts quickly.
  • For journalists and fact-checkers: the Romanian scam pattern — a fake candidate endorsement of a 1,400 lei investment returning 9,000 lei monthly — is likely to be copied in Brazil. Financial-fraud teams should coordinate with election fact-checking desks before the campaign heats up.