Brazil · POLITICS
Key Facts
- —Meeting The TSE held a closed-door meeting on 18 August 2026 to discuss AI rules, but ended without consensus.
- —Current rule Resolution 23,732/2024 requires explicit labeling of AI-generated content and bans deepfakes.
- —Penalty Candidates using deepfakes risk cassation, meaning loss of candidacy or mandate.
- —AI allowed Artificial intelligence is permitted in campaigns if content is labeled as synthetic and technology is identified.
- —Agreement The TSE signed a cooperation deal with ElevenLabs in August 2026 to combat fake synthetic audio.
Ministers met behind closed doors in August, yet the existing ban on deepfakes remains untouched as the 2026 campaign gets underway.
Brazil deepfake rules remain unchanged after the Superior Electoral Court failed to reach an agreement in closed-door meetings this August. The court’s existing resolution, in force since February 2024, continues to ban deepfakes while allowing labeled AI content.
Court Meeting Ends Without Consensus
However, the Superior Electoral Court (TSE) held a closed-door meeting on 18 August 2026 to discuss AI use in campaigns. Brazil deepfake rules ended without agreement.
According to O Globo on 19 August, ministers failed to reach consensus. Therefore, they set no date to discuss the video simulating former President Jair Bolsonaro.
The meeting was called by TSE President Kassio Nunes Marques, who took office in May. Meanwhile, Brazil deepfake rules were a key topic.
The court’s statement called the session ‘quite productive’ but deferred any decision to the plenary. The initial meeting scheduled for 13 August was postponed because the plenary session ran long, yet Brazil deepfake rules remained unresolved.
Nevertheless, the rescheduled gathering on 18 August also failed to produce a consensus. The meeting was held behind closed doors before the plenary session, as Agência Estado reported.
However, the court’s statement did not specify a new date for further discussions.
Brazil deepfake rules Stand Firm
The rules in force come from Resolution 23,732, adopted on 27 February 2024, which amends the 2019 electoral propaganda resolution. In addition, it requires any AI-generated synthetic content to be labeled explicitly, prominently, and accessibly, including naming the technology used.
Moreover, the resolution bans deepfakes. In fact, these are synthetic audio or video that creates, replaces, or alters a real, deceased, or fictitious person’s image or voice.
Even so, consent does not change the definition. This ban applies in any circumstance, whether for negative or positive campaigning.
These rules were established in February 2024 and remain unchanged. Although the court discussed them, no amendments were made in August 2026.
The resolution also requires a watermark or audio description on still images and video. In addition, printed material must include the notice in its body, as practitioner guidance explains.
AI Allowed, Deepfakes Prohibited
The TSE’s position is not a general ban on artificial intelligence, as CNN Brasil reported on 18 August. In fact, candidates may use AI to alter text, video, or audio.
They must state explicitly that the material was artificially made and identify the technology used. In short, the legal position is clear: AI is permitted, but deepfakes are prohibited.
This distinction is central to understanding the court’s approach. This distinction was reaffirmed by CNN Brasil on 19 August, stating that AI use is allowed but deepfakes are prohibited.
Therefore, candidates can use AI as long as they label the content clearly. CNN Brasil reported on 18 August that the TSE’s position is not a general ban.
Meanwhile, the distinction was reaffirmed on 19 August, with AI allowed but deepfakes forbidden.
Penalties for Deepfake Use
Therefore, the resolution warns that a candidate who uses a deepfake may face cassation. This means loss of their candidacy registration or certificate of election if already elected.
Responsibilities are determined under the Electoral Code. Justice Alexandre de Moraes, who does not sit on the TSE bench, issued a public warning on 21 August.
Speaking at a law school, he said that using disinformation mechanisms ‘pumped up’ by AI would lead to cassation. The penalty for using a deepfake is cassation, which means losing the candidacy or elected office.
This warning was reiterated by Justice Moraes, although his remarks were not a formal decision. The penalty of cassation is defined as the loss of candidacy registration or certificate of election.
However, Moraes’s warning was not a formal decision, as he does not sit on the TSE bench.
Moraes Speech Not a Decision
Moraes’s remarks were part of a speech, not a court decision. Although he is not among the ministers who met, appeals can be taken to the Supreme Federal Court.
There, he may eventually deliberate. Earlier, on 31 July, Moraes indicated that using an AI avatar of Jair Bolsonaro could result in cassation of Flávio Bolsonaro’s candidacy.
However, this was at the PL convention. That is if it spreads disinformation that unbalances the contest.
Moraes’s speech at the University of São Paulo was a public warning, not a court ruling. As a result, it does not change the current legal framework.
His remarks were made at a law school, not in a court session. Therefore, they do not constitute a binding ruling on any candidate.
Campaign Context
Meanwhile, the closed-door meetings on 13 and 18 August fell in the same fortnight as the candidate registration deadline on 15 August. They also coincided with the start of general campaigning on 16 August.
So the court was debating AI rules with the campaign already underway. The first round of Brazil’s 2026 general elections is set for 4 October, with a possible run-off on 25 October.
Party conventions ran from 20 July to 5 August. The meetings occurred just after the start of general campaigning on 16 August.
Therefore, the court was deliberating with the campaign already in progress. The meetings occurred just after the start of general campaigning on 16 August.
However, the court was still deliberating on how to apply the AI rules.
No New Rules Adopted
Despite the meetings, no new resolution or amendment was adopted in August 2026. Still, the existing rules from Resolution 23,732/2024 remain exactly as they were on 27 February 2024.
The meetings were aimed at aligning ministers’ understanding and guiding future judgements, not at changing the rules. No decision was made.
The August meetings were for discussion only, not for enacting changes. Consequently, Resolution 23,732/2024 remains the governing rule for AI and deepfakes.
The research found no record of a plenary judgement on a specific deepfake resolution in August. As a result, the existing rules remain unchanged.
Platform Obligations
The resolution also imposes duties on platforms, requiring them to act proactively against disinformation affecting the electoral process. It restricts the use of robots to simulate dialogue with candidates or others.
Big tech companies must remove content with disinformation, hate speech, Nazi and fascist ideology, and anti-democratic, racist, or homophobic material immediately. These obligations remain in force.
Platforms must also remove illegal content immediately, including hate speech and disinformation. These duties are part of the same resolution and were not altered in August.
The resolution also restricts the use of robots to simulate dialogue with candidates or others. In addition, platforms must remove illegal content immediately, including hate speech and disinformation.
Cooperation With ElevenLabs
In short, in August 2026, the TSE signed a technical cooperation agreement with British AI company ElevenLabs. This was to curb fraudulent use of synthetic audio during the elections.
The agreement is part of broader efforts to address AI challenges. Meanwhile, the court has not set a new date to discuss the specific video or any further action.
The campaign continues under the existing rules. The TSE’s agreement with ElevenLabs aims to prevent fraudulent synthetic audio.
However, this cooperation was announced separately from the August meetings. The agreement with ElevenLabs was announced separately from the August meetings.
However, it aims to curb fraudulent synthetic audio during the elections.
Frequently Asked Questions
What are Brazil’s deepfake rules?
Brazil’s deepfake rules are set by Resolution 23,732/2024, which requires labeling of AI-generated content and bans deepfakes outright. The ban applies to synthetic audio or video that alters a person’s image or voice, even with consent.
Did the TSE change the rules in August 2026?
No, the TSE did not change any rules. The closed-door meeting on 18 August ended without consensus, and no new resolution was adopted.
Is artificial intelligence banned in Brazilian elections?
No, artificial intelligence is allowed, but it must be labeled. Candidates can use AI to alter content if they state explicitly that it was artificially made and identify the technology used.
What happens if a candidate uses a deepfake?
A candidate who uses a deepfake may face cassation, meaning the loss of their candidacy or mandate. The penalty is outlined in the TSE resolution, and responsibilities are determined under the Electoral Code.
Connected Coverage
Sources: O Globo; Agência Estado via UOL; Poder360; CNN Brasil; R7; Gazeta do Povo; g1; ConJur; Jusbrasil; Tribunal Superior Eleitoral; Brazilian Chamber of Deputies.
This article was produced by The Rio Times’ automated newsroom system. How we use AI · Report an error
Read More from The Rio Times