Brazil's electoral court ordered an unlabelled AI post by Flávio Bolsonaro removed under Resolution 23.610 art. 9-B

Key takeaways

• The TSE ordered removal of three X posts by Flávio Bolsonaro, unanimously, in Rp 0600955-65.2026.6.00.0000.
• One post was removed for lacking the AI identification that art. 9-B of Resolution TSE 23.610/2019 requires. The other two were removed on a separate ground, the absence of a minimal factual basis.
• The order sets a 24-hour removal period, a daily fine for non-compliance and a bar on republishing substantially identical content.
• The source is a TSE news release. It does not give the fine amount or the full reasoning.

What did the TSE decide about the Flávio Bolsonaro posts?

According to the release, the TSE ordered removal of three posts that Flávio Nantes Bolsonaro published on X on 30 and 31 May 2026. The decision came in a representation filed by the Federação Brasil da Esperança (FE Brasil), led by the Partido dos Trabalhadores (PT), which alleged early negative electoral propaganda against Luiz Inácio Lula da Silva. The release states the vote was unanimous and the rapporteur was Minister André Mendonça.

One post used AI to simulate Lula in a situation that did not exist. The other two, according to the representation, associated the then pre-candidate with criminal organisations, the PCC and the CV. The release records the allegation as the complainant's; the ruling on those two posts is covered below.

What does the ruling say about the missing AI label?

The rapporteur reconsidered an earlier decision that had denied the request for urgent relief. The release says he found the first post had a caricatured and fanciful look, but that the content was not identified as produced with artificial intelligence.

The release then reports his reasoning in these words: "a ausência da identificação exigida pelo artigo 9º-B da Resolução-TSE nº 23.610/2019 configura descumprimento do dever de transparência". In the release's own description, art. 9-B requires political or electoral content fabricated or manipulated with AI to be "identificados de forma expressa, destacada e acessível". The release says the missing label authorises removal of the post.

We read this as a ruling about labelling, not about whether the image was believable. The release does not say the post was deceptive, only that it was unlabelled.

What happened to the two posts about criminal organisations?

These two were handled on a different ground. The release says the rapporteur applied the Tribunal's most recent case law on early negative electoral propaganda. In the understanding applied to the case, "a associação de candidato a organização criminosa, quando não há substrato fático mínimo, ultrapassa os limites da crítica política". The release adds that such an association can amount to serious decontextualisation. This ground does not turn on AI.

The release adds the operative terms. The three posts are to be removed within 24 hours, under pain of a daily fine. Flávio Bolsonaro must also refrain from republishing or spreading the analysed media or substantially identical content, and the provider responsible for the X platform is to be notified so that it can comply.

What does it mean for election lawyers, platforms and campaign counsel?

For election lawyers and campaign counsel, the release shows the TSE treating a missing AI identification as sufficient on its own for a removal order. The release reports that the rapporteur reconsidered an earlier denial of urgent relief; it does not say what procedural stage the case is otherwise at. For platforms, the release says the X provider is to be notified of the order. It does not say what the provider must do beyond complying with the determination, and it does not say whether any platform has been fined.

The release does not discuss whether the ruling binds other cases. Read it as one data point on how art. 9-B is applied, with the full decision still to be checked before relying on it in a filing.

What should campaign counsel and platforms do now?

The release states no compliance steps. The points below are the desk's suggestions drawn from it.

• Ask whether each piece of political content made or altered with AI carries an express, prominent and accessible AI identification, which is how the release describes art. 9-B.
• Do not treat a satirical or exaggerated look as a substitute for the label. The rapporteur, as reported, saw the caricatured appearance and still required identification.
• Our suggestion, not something the release requires: assign one person to confirm the label before posting and keep a dated record of that check.
• Obtain the full decision for Rp 0600955-65.2026.6.00.0000 from the TSE case system before citing it.

What we did not verify

What we opened: the TSE news release of 8 October 2026 (published 14:11, updated 15:44), on 2026-10-25.

What we did not open: the decision itself, the docket entries, the text of Resolution TSE 23.610/2019 and its art. 9-B, the earlier decision denying urgent relief, the posts, and any statement from Flávio Bolsonaro or his counsel. We do not know the fine amount, whether the posts were in fact removed, whether any appeal was filed, or how the 24-hour period is counted.

What we refuse to claim: that the ruling sets a rule for all AI election content, that platforms face any sanction, or any holding beyond what the release reports. Our description of art. 9-B is the release's description, not our reading of the resolution.

Related briefings

• AI Regulation News hub
• Latin America AI regulation tracker
• Attorneys and AI regulation
• TSE Resolution 23.755 on election deepfakes and AI labelling
• STF Justice Moraes order on an AI avatar and electoral deepfakes
• TSE and AGU accord on AI good practices in elections
• ANPD technical note on Grok, X and deepfakes under the LGPD