Brazil's Superior Electoral Court (TSE) has spent two election cycles fighting platforms after the fact — ordering takedowns, issuing fines, negotiating case-by-case agreements. On 27 July 2026 it tried something structurally different: Portaria No. 463 forces platforms to show their homework in advance. Providers with more than 5 million active monthly users in Brazil now have until 16 August 2026 to file compliance plans detailing exactly how they will detect election risks, identify coordinated inauthentic behavior, and stop generative AI tools from tilting toward a candidate. The Portaria implements Article 125-B of the TSE's underlying elections rulebook, Resolução 23.610/2019, as amended by Resolução 23.755 of 2 March 2026, and applies to any provider whose service lets users publish, share, recommend, boost, or monetize political content — plus, notably, messaging apps with public channels and generative-AI tools themselves.
Why the TSE Is Moving First This Time
The timing is not incidental. Brazil votes in October 2026 with former president Jair Bolsonaro barred from the ballot and serving a prison sentence for a coup attempt that prosecutors say included a plot to assassinate political rivals; his son, Senator Flávio Bolsonaro, now carries the family's political weight against incumbent Luiz Inácio Lula da Silva. Two cycles of platform-fueled disinformation, culminating in January 8, 2023's attack on federal buildings, have made the TSE deeply skeptical of platform self-policing. Layered onto that is a documented AI problem: Sumsub's 2025 fraud-monitoring data, reported by Agência Brasil, found AI-manipulated video in Brazil surged sharply year-on-year, and Resolução 23.755 responds directly, requiring conspicuous labeling of AI-generated political content and, under Article 9º-B §3º-A, banning new synthetic content depicting candidates or public figures in the 72 hours before polls close through 24 hours after.
The Case the TSE Is Making
There is a real argument for the compliance-plan model, and it deserves to be stated plainly before it's contested. As Tatiana Dourado, a professor at PUC-Rio, told Tech Policy Press, requiring platforms to "provide complete, detailed, comprehensive information" up front is a shift from reactive firefighting to structural accountability — regulators reviewing a platform's actual detection criteria and escalation flows before an election, rather than negotiating over individual pieces of content once damage is already circulating. Compliance plans that specify false-positive rates, fact-checking partnerships, and audit trails, as Article 4 and Article 7 of the Portaria require, are also harder for a platform to fudge than a one-line pledge to "combat misinformation." And the enforcement lever is genuinely consequential: non-compliant platforms face either direct TSE supervision or a ban on political advertising, cutting off a significant campaign-season revenue stream.
Where the Framework Strains
But good intentions in the design don't guarantee good outcomes in the execution, and this framework has real soft spots. First, capacity: Dourado herself flagged "another challenge on the TSE's side on the capacity to follow and track each compliance plan" — a single court now has to substantively evaluate methodology documents from every major platform operating in Brazil, on a timeline compressed by an election two months out. Political scientist Andressa Michelotti raised a sharper concern to Tech Policy Press: given thin moderation staffing and looming obligations around the US midterms the following month, platforms facing ambiguous, high-stakes 'risk hypothesis' criteria may simply over-remove — "we may end up seeing a scenario where removing is easier than keeping content online." That is precisely the failure mode a pro-speech, evidence-based publication should worry about: a compliance regime that quietly rewards blunt takedowns over precise ones.
Second, the Portaria's own text undercuts its transparency promise in places. Data Privacy Brasil's analysis of the document found that Article 9, §§2–5 lets providers withhold information by citing commercial secrecy or foreign-jurisdiction constraints, without objective criteria distinguishing a legitimate refusal from one that guts oversight — and that AI "resistance test" results under Article 15(V) are restricted even from the TSE's own final public reports, closing off the kind of independent scrutiny that gives compliance regimes credibility in the first place.
Third, the 72-hour pre-election blackout on synthetic content, while aimed squarely at deepfakes, is drawn broadly enough that legal analysts at Migalhas have flagged tension with Brazil's constitutional free-expression guarantees (Article 5, IV and IX) over whether it reaches only manipulative fakes or AI-assisted content generally.
A Proportionate Verdict
None of this means the TSE was wrong to act. A documented deepfake surge and two elections' worth of platform-enabled disinformation are a legitimate basis for regulation, and requiring specific, auditable plans is more proportionate than blanket content mandates or preemptive bans. But proportionality cuts both ways: the same instrument needs narrower risk-detection language, harder limits on the secrecy carve-outs, and public access to the AI safety testing the TSE now keeps sealed. Brazil built a genuinely interesting model here — accountability before the crisis instead of litigation after it. Whether it protects speech as well as it protects the vote will depend on details the TSE has, so far, kept to itself.