A comprehensive bill lands mid-hub-building season
On July 15, 2026, national deputy Juan Fernando Brügge (Córdoba, Provincias Unidas) filed a bill in the Chamber of Deputies establishing a comprehensive legal framework for artificial intelligence in Argentina (bill text, HCDN; Parlamentario). The bill sorts AI systems into three tiers — unacceptable, high, and limited risk — bans social-scoring mechanisms and behavioral-manipulation systems targeting vulnerable people, restricts real-time remote biometric identification to narrow judicial exceptions, and requires human rights impact assessments before high-risk systems (health, employment, education, migration, critical infrastructure, biometric ID) go live. It also creates a carta de derechos: the right to know when you're dealing with an automated system, to understand the criteria behind a decision, and to demand human review when that decision materially affects you. The National Institute of Industrial Technology (INTI) would enforce it, backed by a multidisciplinary advisory council. Committees on science and technology, general legislation, and human rights will review it first.
The timing is pointed. Brügge revives, in the same package, penal and civil sanctions he first proposed in 2024 — after lobbying from the Argentine actors' union — for undisclosed deepfakes and manipulated media, with 4-to-8-year prison terms for malicious fabrication (APFDigital). That bill stalled in 2024; folding it into a broader risk-based statute now is a bet that comprehensive framing succeeds where a narrow one didn't.
The case for it
The strongest argument for Brügge's bill isn't hypothetical. Argentina has no dedicated AI regulator today — only a 2023 non-binding resolution from the data protection authority (AAIP), which created a voluntary transparency and personal-data program for AI grounded in Argentina's existing privacy and access-to-information statutes (Boletín Oficial, Resolución 161/2023). A guidance document with no enforcement teeth doesn't stop a credit-scoring algorithm from encoding discrimination, a hiring tool from opaque rejections, or a biometric system from misidentifying someone at a border checkpoint with no appeal. Automated decisions in health, employment, and migration carry real stakes, and a carta de derechos that guarantees a human review path when they go wrong is a defensible floor, not regulatory overreach. Reasonable people can disagree about the mechanism while agreeing the gap is real.
The case against copying the EU exactly, right now
But Brügge's own account of the bill leans on "international experience, especially European risk-based regulation" — and the EU's risk-based model is, at this exact moment, proving harder to operationalize than its architects expected. On May 7, 2026, EU lawmakers agreed to push back the AI Act's high-risk compliance deadline for Annex III systems (biometrics, employment, credit, law enforcement, migration — the same categories Brügge's bill targets) from August 2, 2026 to December 2, 2027, a 16-month delay, with product-embedded systems pushed to August 2028 (Travers Smith; European Commission). The stated reason: implementation guidance and technical standards weren't ready when the deadline arrived. That's the world's most resourced digital regulator, with a multi-year head start and a dedicated AI Office, still finding that ex-ante licensing-style classification outpaces institutional capacity to administer it. Assigning the same job to INTI — an industrial-standards body with no history of adjudicating human rights impact assessments — without first building that capacity risks reproducing the EU's timeline problem without the EU's fallback resources.
The stakes for getting the sequencing right are unusually concrete for Argentina. In October 2025, OpenAI and Sur Energy signed a letter of intent for a $25 billion, up-to-500-megawatt AI data center under the government's RIGI large-investment incentive regime, explicitly framed as making Argentina a regional AI hub (El Litoral). That deal remains non-binding, but it signals where the executive branch has placed its bet — on inbound compute investment, not a compliance regime modeled on a bloc that just admitted its own version needs more runway.
What proportionate would look like
None of this means Argentina should do nothing. The deepfake provisions and the social-scoring ban are narrow, harm-specific, and don't require standing up a new licensing bureaucracy — they can move on their own track quickly. The harder, comprehensive risk-tier apparatus is where sequencing matters: give AAIP's existing resolution statutory teeth and enforcement funding before layering a second, INTI-run risk-classification regime on top, and calibrate the high-risk compliance runway to what INTI can actually administer rather than importing the EU's original (now-abandoned) two-year clock. A carta de derechos is worth having. It's worth having in a form the enforcing agency can actually deliver on day one.