Argentina spent the last year cycling through deepfake bills almost as quickly as it cycles through finance ministers. The most consequential is Bill 5945-D-2025, introduced in the Chamber of Deputies by Diputada Gisela Marziotta. It proposes a fast judicial channel for victims of AI-generated image abuse. A separate proposal, Expediente 7225-D-2024, would carve new offences into the Criminal Code. Behind both sits the AAIP's earlier soft-law work — Resolution 161/2023 created the Programa de Transparencia y Protección de Datos Personales en el uso de la Inteligencia Artificial, published in the Boletín Oficial on 4 September 2023, but stopped short of binding rules on synthetic media.
The debate sharpened this month. An 11 May 2026 op-ed in Infobae argued Argentina should follow Denmark's emerging "image-as-IP" model, granting individuals exclusive rights over their face and voice. We disagree. The Marziotta bill's court-centred approach is a better fit for Argentine constitutional tradition — and more compatible with free expression — than either Danish-style IP commodification or European-style platform pre-screening mandates.
What the Marziotta bill actually does
The bill grafts a new acción especial onto Articles 51–53 of the Civil and Commercial Code, which already recognise dignity and image rights. The mechanics, per the text filed with the Cámara de Diputados:
- Any affected person — or their representative or heirs — can file digitally before any first-instance judge.
- No court fees. No formal pleading requirements.
- The judge must rule within 24 hours, including weekends.
- Courts may order blocking, de-indexing, or removal as precautionary measures, and must order permanent deletion upon confirming unauthorised use.
- Daily fines, proportional to harm, apply for non-compliance. Platforms face temporary operational suspension as a last resort.
The bill explicitly cites the US Take It Down Act of 2025 as inspiration and distinguishes itself from copyright-based approaches under discussion in Denmark.
The Córdoba case driving the urgency
The strongest argument for the bill is the case the Argentine press is calling the deepfake escolar. In June 2025, a Córdoba prosecutor's office elevated to trial a 19-year-old — identified in court papers as J.M.C. — who had downloaded classmates' photos from social media, face-swapped them into pornographic videos using AI tools, and uploaded the results to adult sites with the victims' real names. Sixteen young women were affected, most minors at the time. Forensic evaluations documented moderate post-traumatic stress symptoms. Prosecutor Pablo Cuenca charged him under the Ley Olimpia gender-violence framework and, separately, for child sexual abuse material.
Existing Civil Code tort remedies could deliver damages — eventually. None of them gets a deepfake off the internet by lunchtime. That is what victims, and a reasonable proportionality test, demand here. Steelmanning the regulators: the harm is concrete, severe, and time-sensitive, and the existing toolkit moves on a timescale measured in years.
Why courts beat ex-ante platform filters
The European temptation is to load duties onto platforms ex ante — content authentication mandates, watermarking obligations, proactive scanning. The Marziotta bill avoids that path, and rightly. Three reasons:
- Ex-ante scanning is over-inclusive. Mandated detection captures parody, journalism, and political speech alongside abuse. The line is hard to draw algorithmically — something even the EU's DSA enforcers concede when they lean on external auditors like AI Forensics to surface concrete harms.
- Ex-ante mandates entrench incumbents. Compliance infrastructure is expensive. Mandatory pre-screening favours the platforms that can already afford trust-and-safety teams; a judicial takedown lane is platform-agnostic.
- The Danish IP model commodifies identity. Treating face and voice as exclusive property creates licensing markets — but it also hands celebrities and influencers more leverage to suppress satire and journalism than ordinary victims gain against actual abuse. Wrong incentives, wrong constituency.
A court-driven, time-bounded amparo focuses state power on the specific moment of harm: a synthetic image circulating without consent. That is more proportionate, and more speech-protective, than building a permanent filter into the network.
Where the bill needs tightening
Pro-innovation does not mean pro-bill-as-drafted. Three concerns deserve serious committee work:
- Ex parte 24-hour orders are blunt. Judges ruling without hearing from the alleged poster, on weekends, with platform suspension on the table, will sometimes get it wrong. The bill should require expedited adversarial review within 72 hours of any precautionary measure — keeping speed for victims while restoring due process for contested cases.
- "Image and personal attributes" is vague. The same elastic language that protects against deepfake porn can capture political memes, deepfake-detection demonstrations, or satire of public figures. The bill should carve out clear safe harbours for journalism, research, and parody — modelled on the Take It Down Act's narrower non-consensual-intimate-imagery focus rather than the broader Danish formulation.
- Platform suspension is disproportionate. A platform losing operation in Argentina because of one non-compliant case is a sledgehammer. Daily fines, escalating predictably, are sufficient leverage.
The bigger picture
Argentina's deepfake debate is a small, instructive corner of the global content-regulation argument. Spain, France, and California are converging on broader regimes. Argentina has, almost by accident, picked the most modest of the proportionate options: keep the substantive rule grounded in the existing Civil Code, accelerate the procedural remedy, and avoid both criminalisation creep and platform-mandate creep. Refined carefully, the Marziotta framework is what proportionate deepfake regulation looks like — and a useful template for the rest of Latin America to study before reaching for heavier instruments.