{"id":137486,"date":"2026-07-31T16:17:36","date_gmt":"2026-07-31T16:17:36","guid":{"rendered":"https:\/\/mailinvest.blog\/index.php\/2026\/07\/31\/brazil-bans-paid-ads-for-deepfakes-in-final-72-hours-before-elections\/"},"modified":"2026-07-31T16:18:39","modified_gmt":"2026-07-31T16:18:39","slug":"brazil-bans-paid-ads-for-deepfakes-in-final-72-hours-before-elections","status":"publish","type":"post","link":"https:\/\/mailinvest.blog\/index.php\/2026\/07\/31\/brazil-bans-paid-ads-for-deepfakes-in-final-72-hours-before-elections\/","title":{"rendered":"Brazil bans paid ads for deepfakes in final 72 hours before elections"},"content":{"rendered":"

<\/a>\n
<\/p>\n

\n

Brazil’s nationwide knowledge safety authority revealed a 163-page technical report on July 29, 2026, mapping how\u00a0deepfakes<\/a> are used to defraud customers, manipulate elections and generate non-consensual intimate imagery, arguing that paid digital promoting has turn into one of many expertise’s most energetic vectors for monetary hurt. The doc, titled “Radar Tecnol\u00f3gico N. 6 – Deepfakes” and issued by the Ag\u00eancia Nacional de Prote\u00e7\u00e3o de Dados (ANPD), was written by a six-person staff led by Ticianne de Gois Ribeiro Darin beneath the company’s Superintend\u00eancia de Inova\u00e7\u00e3o Tecnol\u00f3gica, in accordance with the report.<\/p>\n

The research carries no binding authorized drive. Based on the report, it’s framed as a technical-informative and potential doc that doesn’t represent normative steerage, an administrative determination or a conclusive institutional place on particular instances. Even so, it compiles enforcement figures, court docket instances and cross-border comparisons that give advertisers, platforms and compliance groups a working map of the place legal responsibility is heading, at a second when the\u00a0European Union<\/a>‘s personal AI Act labeling regime turns into legally binding on August 2, 2026, three days after the ANPD’s publication date.<\/p>\n

A fraud sample with a well-known form<\/h2>\n

The report devotes a full part, numbered 5.6, to what it calls misleading promoting via deepfakes, describing a three-step operational sample that recurs throughout the Brazilian instances it paperwork. First, felony teams create artificial content material utilizing the picture and voice of a public determine. Second, that content material is boosted on social networks as sponsored materials focused by consumer profile. Third, victims are redirected to fraudulent environments, cloned web sites or faux platforms, the place they’re requested for funds or private knowledge.<\/p>\n

Based on the report, one of the crucial consequential instances concerned a felony group that used deepfakes of celebrities together with Gisele Bundchen, Angelica, Juliette and Sabrina Sato to promote merchandise that didn’t exist. Within the Bundchen case, the report states, the group offered a supposedly free anti-wrinkle package for which victims paid solely a transport payment via Brazil’s PIX instant-payment system; the product was by no means delivered. Police investigation discovered the group had moved greater than R$20 million, and its members had been arrested, the report says, citing a 2025 information account.<\/p>\n

A second advertising-fraud sample documented within the report entails the repeated manipulation of well being professionals’ likenesses. Doctor Drauzio Varella’s picture and voice, the report notes, had been manipulated on a number of events to simulate endorsements of medicines and coverings with out scientific backing, usually introduced in a format resembling interviews or information segments to borrow the credibility of journalism. The report contains an illustrative determine contrasting a fabricated account, recognized as @dr.mauro_corella_tv, towards the doctor’s real, verified account discussing a authentic public-health matter.<\/p>\n

The report characterizes the ensuing hurt as one outlined by scale reasonably than dimension: every sufferer tends to lose a comparatively small quantity, which reduces the chance of a proper criticism whereas permitting operators to build up substantial combination revenue, citing Romagna’s 2025 evaluation. Past direct monetary loss, the report lists 4 secondary results: danger to public well being the place unproven medical merchandise are concerned; harm to the credibility of execs whose likenesses are used with out consent; violations of persona rights; and publicity of private knowledge, since victims usually submit data on to fraudulent web sites.<\/p>\n

Monetary impersonation past promoting<\/h2>\n

A separate part of the report, numbered 5.5, catalogs three further Brazilian financial-fraud instances involving deepfakes that sit outdoors standard promoting however illustrate the identical underlying danger. One facilities on a video utilizing the picture and speech of Pedro Moreira Salles, co-chairman of the board of Banco Itau, shared extensively on Fb in 2025 and attributing false statements to the chief to advertise a supposed miracle funding scheme; Itau subsequently clarified he had no connection to the fabric, in accordance with the report. One other, the report says, concerned a Rio Grande do Sul lady who believed she was in a digital relationship with actor Brad Pitt, having carried out what she described as video calls with an artificial model of the actor generated in actual time; the identical fraud sample value a lady in France greater than 830,000 euros, prompting the actor’s consultant to difficulty a public warning about unauthorized use of his likeness.<\/p>\n

A 3rd case considerations Serasa Experian, the Brazilian credit-data firm. Based on Could 2026 knowledge cited within the report, roughly 83.5 million folks in Brazil, near half the grownup inhabitants, carry detrimental credit score information, whereas an estimated 57 million customers maintain money owed they’re unaware of, a inhabitants the report calls particularly weak to scams exploiting the Serasa identify. Fraudulent content material promising compensation of as much as R$30,000 for a supposed 2021 knowledge leak circulated broadly, in some instances utilizing a deepfake video of then-senator Sergio Moro, who later publicly clarified it was false. No court docket has but ordered Serasa Experian to pay such compensation, the report notes, although the underlying 2021 leak, which affected roughly 223 million folks, is now the topic of a category motion in English courts towards the Experian group.<\/p>\n

To quantify the broader development, the report cites the Identification Fraud Report 2025-2026 from identity-verification firm Sumsub, discovering Brazil accounted for roughly 39 % of detected deepfake instances in Latin America throughout 2025, alongside a 126 % enhance in deepfake and synthetic-identity assaults between 2024 and 2025.<\/p>\n

A number of of the fraud instances the report paperwork had been disseminated particularly via Fb, a element that locations the ANPD’s findings alongside enforcement motion Meta has already taken in the identical market.\u00a0Meta<\/a>\u00a0filed multiple lawsuits on February 26, 2026 against deceptive advertisers operating in Brazil, China and Vietnam<\/a>, concentrating on operators who used celeb impersonation and cloaking methods practically similar to the sample the ANPD report describes: artificial endorsements paired with redirection to fraudulent websites. Meta eliminated greater than 134 million rip-off ads throughout 2025, in accordance with disclosures made at a International Anti-Rip-off Summit in Washington, and expanded facial-recognition expertise for celebrity-impersonation detection, which it stated greater than doubled fraudulent-ad detection throughout testing. Inner Meta paperwork reported by Reuters in November 2025 projected that roughly 10 % of the corporate’s 2024 income, roughly $16 billion, got here from ads it internally categorized as selling scams, a characterization Meta has disputed; a\u00a0Consumer Federation of America class action filed against Meta on April 26, 2026<\/a>\u00a0argued the platform’s auction-based pricing rewarded lax fraud enforcement. Individually, the\u00a0Federal Trade Commission found that social media scams cost Americans $2.1 billion in 2025<\/a>, and newer FBI knowledge attributed $893 million in reported 2025 losses particularly to AI-linked fraud.<\/p>\n

Different platforms have moved on associated enforcement.\u00a0Google’s AI-powered defense suspended 39 million advertiser accounts<\/a>\u00a0after assembling a staff of greater than 100 specialists to counter deepfake celebrity-endorsement scams, completely suspending greater than 700,000 offending accounts.\u00a0Microsoft Advertising revised its policies in October 2024 specifically to address deepfake technology<\/a>\u00a0and fraudulent endorsements, later reporting elimination or restriction of a couple of billion ads throughout 2024.<\/p>\n

Brazil’s electoral court docket strikes towards paid artificial content material<\/h2>\n

Outdoors the advertising-fraud context, the ANPD report devotes its longest regulatory dialogue to Brazil’s Tribunal Superior Eleitoral (TSE), the nation’s electoral court docket, which the report credit with taking essentially the most concrete regulatory motion towards deepfakes of any Brazilian establishment to this point. Below Decision 23.732\/2024, which amended an earlier 2019 decision, the TSE prohibited using artificial audio or video content material, nevertheless generated, to create, exchange or alter the picture or voice of a residing, deceased or fictional individual to profit or hurt any candidacy, in accordance with the report. The identical decision requires that artificial content material utilized in electoral promoting carry an specific, outstanding disclosure figuring out it as fabricated and naming the expertise used, the report states.<\/p>\n

In 2026, the report notes, the TSE went additional with Decision 23.755\/2026, which prohibits the publication, republication, even the place unpaid, and particularly the paid boosting of recent artificial content material depicting a candidate or public determine’s picture, voice or likeness through the 72 hours previous and the 24 hours following the shut of polling, no matter whether or not that content material carries a disclosure label. This closes a spot the sooner decision left open, in accordance with the report: labeling compliance is now not enough to allow paid distribution of artificial political content material in essentially the most electorally delicate window.<\/p>\n

This element locations Brazil’s electoral regulator forward of the transparency-only mannequin that has outlined most deepfake-specific promoting regulation elsewhere. The report positions Brazil’s strategy inside three broad worldwide patterns it identifies: transparency mandates requiring disclosure of artificial content material; electoral prohibitions limiting deepfake use round voting; and criminalization of essentially the most damaging classes of abuse, mainly non-consensual sexual content material. South Korea has banned deepfake marketing campaign movies for 90 days earlier than an election beneath its Public Official Election Act, the report notes, whereas 28 of fifty US states handed legal guidelines between 2019 and 2025 requiring detection and labeling of artificial election content material, led by California and Texas.<\/p>\n

The labeling commonplace the European Union has already chosen<\/h2>\n

The transparency class the report describes maps intently onto the regime advertisers within the European Union are actually confronting straight. Based on the report, Article 50 of the EU’s\u00a0AI Act<\/a>\u00a0requires suppliers of generative AI programs to make sure artificial audio, picture, video or textual content output is marked and detectable as artificially generated, whereas individually requiring anybody deploying a system that generates deepfake-constituting content material to reveal that it was artificially generated or manipulated. The report additionally notes that Article 50 exempts instruments used purely for routine modifying that don’t alter a bit of content material’s precise which means, distinguishing deepfakes from odd filters or shade correction.<\/p>\n

That regime is now not theoretical for the promoting business.\u00a0Google signed the EU’s Code of Practice on Transparency of AI-Generated Content<\/a>\u00a0as Article 50 obligations grew to become relevant, with non-compliance carrying fines of as much as 15 million euros or 3 % of worldwide annual turnover, whichever is greater.\u00a0Meta signed the same code five days before the deadline<\/a>, with the corporate’s VP of Public Coverage for Europe, Markus Reinisch, stating in a newsroom put up that signing “displays our dedication to this precept,” referring to giving folks instruments to establish artificial media. The\u00a0European Commission’s implementation guidelines, published July 20, 2026<\/a>, explicitly exclude persuasive business content material from the lighter disclosure regime reserved for creative or satirical work, a studying that straight impacts entrepreneurs utilizing AI-generated spokespeople or avatars in campaigns.<\/p>\n

On the advertiser-facing aspect,\u00a0Google Ads API version 24.2, released June 24, 2026<\/a>, launched developer-facing synthetic-content labeling buildings masking two dimensions: whether or not an asset or advert is AI-generated, and whether or not it was produced totally routinely or with advertiser assessment. A separate consumer-facing “How this advert was made” panel adopted on July 9, 2026, extending into Google’s advert merchandise a disclosure logic already utilized on YouTube, the place\u00a0C2PA metadata and SynthID watermarks trigger automatic content labels<\/a>\u00a0even when a creator doesn’t choose the disclosure possibility manually throughout add.<\/p>\n

The ANPD report devotes its remaining substantive part to technical mitigation instruments, describing 4 classes in use: automated detection primarily based on algorithms educated to identify manipulated-content patterns; cryptographic signatures verifying content material integrity from origin; digital watermarking, utilizing seen or invisible alerts figuring out AI-produced content material; and traceability requirements, citing the\u00a0Coalition for Content Provenance and Authenticity<\/a>, or C2PA, which the report describes as recording a digital file’s origin, authorship and edit historical past.<\/p>\n

Even so, the report is candid in regards to the limitations of at the moment out there detection software program. It cites a comparative 2025 research by Moraes and Batista inspecting three business instruments, DeepWare Scanner, Attestiv and TrueMedia, which discovered that none reliably distinguished deepfakes from genuine materials, together with low-resolution footage or content material produced beneath tightly managed circumstances equivalent to archival tv clips. All three instruments returned unsure outcomes for each artificial and real materials, main the research’s authors to conclude that combining automated instruments with human analysis stays important for sturdy verification. The report frames this as proof that detection expertise alone can not substitute for a broader governance strategy that additionally contains regulation, digital literacy and traceability infrastructure.<\/p>\n

Brazil’s personal AI decree and enforcement backdrop<\/h2>\n

The report situates its advertising-fraud findings inside Brazil’s wider 2026 AI regulatory exercise. Decree 12.976, signed Could 20, 2026, prohibits web utility suppliers from producing or modifying intimate content material depicting third events utilizing synthetic intelligence, a measure aimed toward non-consensual sexual deepfakes reasonably than business fraud, although the report treats it as a part of the identical regulatory momentum. The ANPD itself expanded its enforcement posture over the identical interval, having\u00a0added child protection to its 2026-2027 enforcement priorities on December 22, 2025<\/a>\u00a0and\u00a0released a draft age-verification guide for public consultation on May 23, 2026<\/a>, each overlapping with the report’s dialogue of AI-generated baby sexual abuse materials as a definite hurt class.Brazil at the moment has no particular felony offense masking abusive deepfake use typically, in accordance with the report; non-consensual intimate deepfakes are as a substitute prosecuted beneath article 218-C of the Penal Code, a 2018 provision whose penalty was elevated in 2025. A separate 2025 regulation, Lei 15.123, amended article 147-B of the Penal Code masking psychological violence towards ladies to extend penalties when AI is used to change a sufferer’s picture or voice, citing UN Ladies knowledge suggesting round 41 % of ladies in public positions report self-censoring to keep away from additional deepfake-enabled assaults.<\/p>\n

Why this issues for promoting and advertising professionals<\/h2>\n

For advertising and promoting professionals, the ANPD report features much less as a brand new authorized instrument and extra as a structured proof base that can probably inform how Brazilian regulators, courts and platforms deal with synthetic-media promoting fraud going ahead. PPC Land’s personal protection paperwork a sample the report independently confirms from the Brazilian aspect: Meta, Google and Microsoft have every constructed particular enforcement infrastructure round deepfake-based celeb endorsement fraud, exactly the mechanism the ANPD identifies because the dominant type of misleading promoting involving artificial media in Brazil. That convergence between a nationwide regulator’s research and platform-level enforcement motion, filed independently and roughly six months earlier by Meta towards Brazilian defendants particularly, suggests regulators and platforms are working from a broadly appropriate image of how this fraud operates, even the place their cures differ, one counting on litigation and account suspension, the opposite on data-protection frameworks that stay largely untested in Brazilian courts.<\/p>\n

The report’s TSE part carries a definite implication for political promoting: a disclosure label, enough beneath the AI Act’s Article 50 and beneath most transparency-style state legal guidelines in the US, just isn’t enough beneath Brazil’s 2026 electoral rule throughout essentially the most delicate pre- and post-election window, the place paid boosting of artificial candidate content material is banned outright no matter labeling. Any promoting or campaign-technology agency working in Brazilian electoral contexts would want to deal with that 72-hour to 24-hour window as an absolute restriction reasonably than a disclosure obligation. That distinction sits alongside a extra uncomfortable discovering: none of three business detection instruments the report examined may persistently distinguish artificial from genuine media, at the same time as platforms more and more depend on related automated detection, paired with C2PA metadata and watermarking, to energy the disclosure programs that labeling regimes now require.<\/p>\n

Timeline<\/h2>\n