Mendonça Offers Deepfake Test as Brazil Security PEC Advances
Brazil · JUSTICE
Key Facts
- —Court the proposal was made at the Superior Electoral Court (TSE), where Mendonça serves as vice-president
- —Test deepfake status would require realism capable of producing a false perception of authenticity
- —Pending case no date is set for judging the AI video of Jair Bolsonaro shown at the PL convention
- —Senate rapporteur Rogério Carvalho will keep the Chamber’s security PEC text intact
- —Vote math the PEC passed the Chamber 487-15 and now needs 49 of 81 Senate votes in two rounds
A restrictive definition of deepfake could help Flávio Bolsonaro’s defence, while the security amendment gets a fast track in the Senate.
Justice André Mendonça on Tuesday 25 August 2026 proposed a formal legal test for deepfakes at Brazil’s Superior Electoral Court (TSE), arguing that only synthetic content realistic enough to create a false perception of authenticity should fall under the ban, a thesis that could reshape Brazil deepfake rules and influence the still-unresolved case over an artificial-intelligence video of Jair Bolsonaro. On the same day in the Senate, the rapporteur of the public-security constitutional amendment, the PEC da Segurança Pública, said he will not alter the text approved by the Chamber of Deputies.
The legal test proposed by Mendonça
Under the thesis laid out by Mendonça, a justice of the Supreme Federal Court who sits as vice-president of the TSE, only synthetic audio or video whose degree of realism or verisimilitude can lead a voter to believe they are seeing an authentic record would be treated as a deepfake. The stated aim of Brazil deepfake rules, in his reading, is to separate potentially misleading material from memes, caricatures, animations and other obviously fictional content produced with artificial-intelligence tools.
The test appeared not in the Bolsonaro case but in a decision ordering the removal of another AI video, published on 14 August by the satirical profile Brasil Sátira do Poder, which placed presidential candidate and senator Flávio Bolsonaro (PL) alongside banker Daniel Vorcaro in fabricated scenes suggesting illicit payments. Mendonça ordered the video removed on 20 August, finding on a preliminary analysis that its photorealistic images could pass as records of real events. He added that simply labelling a piece as AI-made does not make lawful a content that otherwise has the characteristics of a deepfake.
Where Brazil deepfake rules stand now
The regulation in force is broader than Mendonça’s proposal. Article 9º-C of the TSE’s electoral-propaganda resolution prohibits, in order to favour or harm a candidacy, synthetic audio or video generated or digitally manipulated to create, replace or alter the image or voice of a living, dead or fictitious person, even with authorization. Breaches can bring fines, removal of content and, in the gravest cases, loss of mandate or registration. The Rio Times has followed this dispute in its coverage of the Bolsonaro deepfake case and of the court’s earlier failure to agree on the rules.
How much support the new test commands inside the seven-member court is unclear. Reporting earlier in August pointed in the opposite direction: several justices then favoured a broad restriction on AI-generated content, while court president Kassio Nunes Marques defended a more flexible interpretation. Mendonça’s thesis has not been put to a vote, and for now deepfake disputes are being settled case by case through individual justices’ rulings rather than a unified standard for Brazil deepfake rules.
The Bolsonaro AI video still awaiting judgment
The case holding the court’s attention dates from 25 July, when the PL’s national convention that launched Flávio Bolsonaro’s presidential bid screened a video recreating his father, Jair Bolsonaro, by artificial intelligence. The former president, convicted for the coup attempt and under house arrest, is barred by Justice Alexandre de Moraes from making political statements during the campaign. The video opened with a warning that the image and voice were artificially produced, which the defence says removes any intent to deceive; the Workers’ Party filed a representation arguing the opposite.
The TSE planned in August to use the case to set the limits of deepfake in the campaign, and the justices discussed a possible ban in mid-August, but no judgment date has been set. With the matter stalled before rapporteur Nunes Marques, related disputes keep arriving at the court and being decided piecemeal, one reason Mendonça’s test, formulated in a separate case, is being read as an attempt to shape the eventual Brazil deepfake rules debate before the convention video is judged.
Senate rapporteur locks in the security PEC text
In the Senate, Senator Rogério Carvalho (PT-SE), rapporteur of the PEC da Segurança Pública in the Constitution and Justice Committee, announced on Tuesday that he will present a favourable opinion without proposing any change to the text approved by the Chamber. Altering the proposal now, he argued, would mean postponing urgent answers that Brazilian society expects in confronting organized crime, and any amendment would force the text back to the lower house for two fresh votes.
The strategy mirrors the one used in past constitutional amendments: if the Senate merely ratifies the Chamber’s wording, the PEC can be enacted without further delay. The government, which authored the original proposal, wants the amendment concluded this year, and Carvalho said his report will recommend immediate approval so the matter can advance to the Senate floor.
What the security amendment would change
Approved by the Chamber on 4 March by 487 votes to 15, after a proposed referendum on lowering the age of criminal responsibility was withdrawn from the text, the PEC constitutionalizes the Unified Public Security System and its financing funds, tightens the legal regime for high-dangerousness criminal organizations, expands the roles of the Federal Police and the Federal Highway Police, authorizes municipalities to create municipal police forces with local and community ostensive policing and adds protections for women, children and adolescents.
For members and leaders of criminal factions, the text mandates harsher penalties scaled to hierarchy, blocks progression to more lenient prison regimes, bars temporary releases and orders the expropriation of assets derived from criminal activity. The proposal now needs the support of at least 49 of the 81 senators in two rounds of voting. Because it is a constitutional amendment, it does not require presidential sanction once approved by both houses.
Frequently Asked Questions
What test did André Mendonça propose for deepfakes?
He proposed that only synthetic content with enough realism to create in the voter a false perception of authenticity should be treated as a deepfake. The test would narrow Brazil deepfake rules by excluding memes, caricatures and clearly fictional AI content from the ban.
Which court is handling the deepfake cases?
The Superior Electoral Court (TSE), which regulates electoral propaganda and enforces Brazil deepfake rules during campaigns. Mendonça is a Supreme Court justice who also serves as vice-president of the TSE, and the pending case over the AI video of Jair Bolsonaro is with the electoral court, not the STF.
What happens next with the security constitutional amendment?
Rapporteur Rogério Carvalho will present a favourable report in the Senate’s Constitution and Justice Committee without changing the Chamber’s text. The PEC then needs 49 of 81 Senate votes in two rounds and, once approved, takes effect without presidential sanction.
Connected Coverage
Brazil’s Top Electoral Court Weighs a Bolsonaro Deepfake — and AI’s Limits
Sources
- www.diariodocentrodomundo.com.br
- g1.globo.com
- www12.senado.leg.br
- valor.globo.com
- www.infomoney.com.br
This article was produced by The Rio Times’ automated newsroom system. How we use AI · Report an error
Read More from The Rio Times