Argentina - AI Regulation Amendment (1472-D-2023)

Overview

Expediente 1472‑D‑2023 proposes targeted amendments to Law No. 25.467 (Science, Technology and Innovation) to explicitly include systems of Artificial Intelligence (AI) within the national science and technology regulatory framework. The bill adds a new article 3° bis establishing core values (including diversity, inclusion, peace and justice) and new articles 8° bis and 8° ter that create (a) a mandatory registration regime for AI systems with the Gabinete Científico y Tecnológico (GACTEC) / supported by the Secretaría para la Tecnología, la Ciencia y la Innovación Productiva (SETCIP), and (b) a complaints and review procedure so any person or legal entity claiming harm or threat from an AI system can request a review. The bill also grants the authority power to order suspension of research, development or deployment when ethical principles are considered to be violated, subject to transparent and appealable procedures. The official project text is available in the Chamber of Deputies record and the project PDF (see HCDN project page and project PDF).

Definitions

The bill does not provide an exhaustive technical glossary but operates with several core definitions implied by the amendments: "AI systems" refers broadly to algorithmic systems, models or software that perform tasks using data and automated decision‑making across their lifecycle; "cycle of life" contemplates stages from research and development through deployment and maintenance; "authority of application" refers to GACTEC, with SETCIP as executive support; "person or legal entity" covers any natural or juridical subject that may be affected by an AI system. Because the bill amends Law 25.467 rather than providing standalone definitions, implementation instruments will be required to produce operational definitions (e.g., what qualifies as an AI system for registration, thresholds for human‑in‑the‑loop requirements, and scope of the "research" vs "application" distinction).

Governance and Institutional Framework

The proposal embeds AI oversight within the institutional architecture already established by Law 25.467: the Gabinete Científico y Tecnológico (GACTEC) — created by prior decrees and responsible for national science and technology policy — is designated the registry and oversight authority, with the Secretaría para la Tecnología, la Ciencia y la Innovación Productiva (SETCIP) acting as the operational secretariat. This leverages existing bodies such as the Consejo Federal de Ciencia, Tecnología e Innovación (COFECyT) and the Comisión Asesora para el Plan Nacional (CAPLANCYT). Implementation therefore anticipates ministerial and interministerial coordination via Argentina.gob.ar - Law 25.467 provisions and the GACTEC/SETCIP connection (see CAPLANCYT). The bill delegates operational rule‑making — registration forms, review procedures, suspension mechanics and appeal channels — to the authority; this model centralizes strategic oversight but will require procedural safeguards and technical capacity building at SETCIP/GACTEC level.

Key Focus Areas

The text concentrates on ethical governance, registration and an administrative review path. Ethical governance is expressed as binding statutory values (diversity, inclusion, peace, justice, non‑subordination of human autonomy, protection of individual/collective security, and environmental coexistence). Registration targets "AI systems in any stage of their cycle of life" to create visibility and traceability. The authority may "order to stop" research, development or application where violations occur, but procedural constraints are stipulated: decisions must be transparent, objective, demonstrably grounded in evidence of harm/threat, and include appeals and protections for researchers and entities. The complaint pathway enables affected parties to request review by SETCIP, creating a public administrative avenue for grievances. The bill prioritizes ethical principles and administrative oversight rather than prescriptive technical standards, leaving room for secondary regulation to set risk tiers, conformity assessment, testing and security obligations. It expressly references international ethical recommendations (e.g., UNESCO) and situates Argentina as seeking leadership in Latin America on ethical AI governance.

Implementation Framework

Operationalization will require the authority to issue implementing regulations and procedural instruments: a public registry platform; registration criteria (scope, required documentation, responsible legal entity, data provenance statements); timing triggers for registration (research funding, public procurement, commercial deployment); confidentiality and IP safeguards balanced against transparency; procedures for evidence collection and risk assessment supporting suspension orders; and an internal appeals mechanism. SETCIP is the logical administrative operator for intake and triage of complaints, while GACTEC — as interministerial cabinet — would exercise higher‑level policy determinations and cross‑sector decisions. The bill’s delegation model makes successful implementation contingent on resource allocation, technical staffing (risk assessors, data scientists, legal analysts), clear service levels and interoperability with existing national registries and sectoral regulators (health, finance, justice).

Monitoring and Evaluation

Monitoring under the draft would combine registry analytics, complaint tracking and periodic reporting to Congress or GACTEC. SETCIP should publish aggregated metrics: number of registered systems, sectors, types of complaints, suspension orders, remediation outcomes, and time to resolution. The law’s emphasis on transparency and objective procedures implies public reporting obligations, although the bill text does not specify periodic frequency or reporting formats. Independent evaluation channels (academic audits, civil society oversight) and alignment with national research evaluation (CAPLANCYT/COFECyT) will be critical to measure whether registration and review lead to reduced harms and improved ethical compliance without unduly restricting legitimate research and innovation.

Penalties, Liability, and Appeals

The bill grants administrative powers (registration enforcement and suspension) and a complaints review route but does not enumerate specific fines or criminal penalties. It requires that any suspension order be transparent, objective and demonstrably tied to risk, and mandates appellate instances to protect rights of researchers and entities. In practice, implementing regulations will need to define graduated sanctions (warnings, mandatory remediation plans, temporary suspensions, prohibition from public funding or procurement, administrative fines) and clarify civil‑liability interactions with existing tort and consumer protection regimes. The absence of explicit punitive scales in the text highlights a design choice to leave sanctions to secondary regulation, increasing the importance of clear, rights‑protective appeal processes in the administrative code.

Relationship to Other Instruments

The amendment directly modifies Law 25.467 and therefore interacts with the existing national science and technology system statutes and decrees that created GACTEC and SETCIP. It complements sectoral rules (healthcare regulation, financial services, data protection laws like Argentina’s data protection regime), and must be implemented in coordination with sectoral regulators where AI systems produce sector‑specific risks. The bill references international ethical guidance (UNESCO) and anticipates that implementing rules will need to align with consumer protection, privacy, and public procurement rules. Because the bill situates AI within the national science system, it also intersects with research funding, academic governance, and institutional ethics review processes.

International Alignment

The bill references international instruments (e.g., UNESCO recommendations) and aligns with global trends favoring ethical AI principles, registries and risk‑based oversight rather than rigid bans. By embedding AI governance in the national science framework, Argentina seeks to harmonize with multilateral guidance while retaining national policy control through GACTEC. The proposed design enables cooperation with international standardization bodies, adoption of best practices on transparency, privacy, and human rights impact assessments, and potential bilateral/regional data governance arrangements. International alignment will depend on secondary regulations addressing conformity assessment, cross‑border data flows, and equivalence of foreign compliance regimes.

Implementation Timeline

Sources and References