Bairo GonzalezLeandro · Martinez
Start your guided journey The story

In the world · Artificial intelligence · 2 March 2026

Deepfakes in Brazil's 2026 elections: the TSE rules

For 2026, Brazil's TSE required labels on AI campaign content, banned deepfakes and barred new AI content 72 hours before the vote. How to stay safe.

bairogonzalez.com team, drawing on Bairo's story · Published

For Brazil's 2026 general elections, the Superior Electoral Court (TSE) approved in March specific rules for artificial intelligence in campaigns: all campaign content made or altered by AI must carry a clear notice, deepfakes are prohibited, AI systems cannot recommend candidates and, in the 72 hours before the vote and the 24 hours after, publishing new AI content with candidates' image or voice is prohibited. In September, the TSE established what counts as a deepfake for the purposes of this prohibition.

What happened

In early March 2026, together with the electoral calendar, the TSE approved the campaign advertising rules for 2026. The summary published by Senado Verifica, the Federal Senate's fact-checking service, highlights:

  • Mandatory label. Electoral content generated or altered by AI must carry a notice that is "clear, visible and easy to understand".
  • Quiet window. In the 72 hours before the election and the 24 hours after, posting new AI-made content with candidates' voice or image is prohibited.
  • No algorithmic recommendation. Platforms cannot use AI to rank, recommend or favor candidates and parties.
  • Platform responsibility. Companies must have compliance plans and reporting channels, and must remove false content.

In September, ruling on a specific case, the TSE established a legal thesis, by majority. According to coverage by Estado de Minas, a deepfake is synthetic content, made or manipulated by AI, realistic enough to simulate the image, voice or statement of a person, living, dead or fictional. The prohibition applies when, in addition to that realism, the material is electoral advertising.

The first round is scheduled for October 4, 2026.

Why it matters

Deepfakes changed the kind of lie. Before, fake news was a text. Now it can be a video with the right face and the right voice, saying what the person never said. And it arrives on the phone, forwarded by someone you trust.

The TSE rules help, but they do not replace the care of the person receiving it. A few simple habits protect you:

  • be wary of video or audio that triggers immediate anger or fear;
  • look for the same information in news outlets and from the TSE itself;
  • notice when the AI label is missing from campaign content;
  • do not forward before checking, especially in the final stretch.

In Bairo's view

For Bairo Leandro Gonzalez Martinez, the deepfake is the opposite of the idea he defends. If AI is an extension of the human, it should extend a person's truth, not fabricate a person who does not exist. In his reading, the problem is not the tool, it is whoever uses it to deceive, and the answer involves two things: clear rules and ways to prove what is true.

He gives weight to the second. For Bairo, at a time when any face can be imitated, proof that there is a real person behind an interaction becomes basic infrastructure. That is why, when describing BTZ Chain, he speaks of a consensus of truth against the fake, the scam and the fraud.

Niche AIs come in as allies of fact-checking: agents that check sources and try to knock down a claim before it goes further. That is the design of CLAIN, the agent orchestrator he founded, in which some research, others check and a person decides. The goal of 142 niche AIs in 142 days, announced on October 1, 2026, is a goal.

Where this meets the ecosystem

None of this is a partisan position. It is about every voter's right to know what is true.

Sources

Read also

More on this topic

In the world: Artificial intelligence