On June 3, 2026, the 2ª Câmara Cível of the Tribunal de Justiça de Mato Grosso do Sul (TJMS) unanimously reversed a lower-court decision and ordered a social media platform to pay R$5,000 in moral damages to a woman from Camapuã whose photograph had been used to fabricate nude images through artificial intelligence. The images, posted to an adult-content profile with degrading captions, drew thousands of views before the platform acted. Desembargador Nélio Stábile, writing for the chamber, held that the fact the nudity was synthetic did not lessen the harm or the platform's duty to act.
Why Article 21, Not Article 19
Brazil's Marco Civil da Internet (Lei 12.965/2014) generally shields platforms from liability for third-party content absent a prior judicial takedown order — that is Article 19's default rule. But Article 21 carves out a narrower, stricter regime specifically for non-consensual intimate imagery: once a victim notifies the platform with enough specificity to locate the content, the platform must act "diligently" to remove it, no court order required. The victim in this case notified the platform on January 21, 2025, through its own reporting channels. The company did not act until she filed suit.
The legal question the case turned on was whether Article 21's protections cover images that are entirely fabricated — no real nudity occurred, no real intimate act was recorded — but depict a real, identifiable person. The chamber said yes: "the use of the Author's genuine photograph to fabricate fake nudity and present it to the public as authentic intimate content reproduces, with particular gravity, the injury that Article 21 seeks to prevent," per the ruling as reported by Correio do Estado and Migalhas. Because the harm is reputational and dignitary rather than physical, the court also found moral damages presumed — the victim did not need to prove psychological injury with evidence beyond the images' existence and reach.
The Case for the Court's Rule
There is a real argument for this outcome, and it deserves to be stated plainly before any pushback. A victim of a fabricated nude image suffers substantially the same reputational, social, and psychological harm as a victim of a leaked real image — the viewers who share and mock the content rarely pause to verify authenticity, and by the time a correction circulates, the damage to the subject's standing is done. If platforms could argue that Article 21's stricter notice regime applies only to genuine footage, that would hand bad actors a functional loophole: fabricate the image rather than steal it, and the platform's takedown duty softens. Courts closing that gap by reading "nudity" to include AI-fabricated nudity is not judicial invention — it is a narrow, textually defensible extension of a regime Brazil already legislated in 2014, long before generative AI made this fabrication trivial and cheap.
Where Proportionality Should Bite
The concern for platforms and speech is not this ruling in isolation — Article 21 already requires a notice specific enough to identify the exact content, which limits it to genuine victim complaints rather than blanket monitoring mandates. The bigger proportionality question sits one article over. On June 26, 2025, the Supremo Tribunal Federal ruled in the joined Temas 987 and 533 that Article 19's general judicial-order requirement is partially unconstitutional, holding platforms liable for a list of grave crimes if they fail to act "imediatamente" — without a court order — and imposing new obligations including local representatives, annual transparency reports, and self-regulatory notice systems, according to ConJur's coverage of the decision. That is a much larger expansion of platform monitoring duty than the TJMS ruling, and it is being built case-by-case through Supreme Court interpretation rather than clear statutory text — a path that leaves platforms guessing at the boundaries of "immediate" action across dozens of content categories, with obvious risk of defensive over-removal of lawful speech.
The Missing Piece
The more durable fix is legislative, not judicial. PL 2338/2023, Brazil's proposed AI regulatory framework, passed the Senate unanimously on December 10, 2024, and has been with the Chamber of Deputies since March 2025, per the bill's official Senado Federal tracking page. Among its provisions are labeling requirements for synthetic audiovisual content and civil-liability rules — including burden-shifting toward plaintiffs who cannot access the technical details of a high-risk AI system — that would give platforms and victims a clear statutory rulebook for deepfakes, rather than forcing every case through tort litigation and case-by-case appellate reasoning. Until that bill clears the Chamber, courts will keep doing this work one lawsuit at a time. The TJMS decision is a sound, narrow application of existing law to a foreseeable new harm. It should not be mistaken for the comprehensive deepfake framework Brazil still needs Congress to pass.