Argentina - AI Program Creation (14/25)

Overview

Resolución PGN 14/25 (Buenos Aires, 14 May 2025) formally creates the Programa de Inteligencia Artificial for the Ministerio Público Fiscal (MPF). The text explains the institutional rationale—rising volumes of digital information, prior digitization efforts (Coirón, SINOPE) and the opportunity to improve operational efficiency—while stressing legal, ethical and technical constraints. The Program is charged with designing strategy, issuing technical and ethical standards, receiving and evaluating project proposals and implementing monitoring, training and inter-institutional cooperation. The resolution expressly draws on international instruments and domestic guidelines, including the UNESCO ethics recommendation and the EU and Council of Europe frameworks. The full text is available from the MPF archive (MPF – Resolución PGN 14/25 (PDF)) and the MPF news page (MPF news – IA institutional survey).

Definitions

The resolution adopts working definitions aligned with major international instruments: "intelligence artificial" is defined as technologies and algorithmic/mathematical models that process data to produce inferences, predictions, recommendations, decisions or content that may affect physical or digital environments with varying levels of autonomy and adaptability. It distinguishes generative AI platforms (text, image or video generators) for which additional traceability and authenticity concerns apply, and situates AI as including both statistical/ML models and more complex adaptive systems. The Program is defined as a cross-cutting institutional structure (not an external regulatory agency) tasked with coordination, standard-setting, evaluation and capacity-building. Key operational concepts used throughout the instrument are: "control human significant" (meaningful human oversight), "risk-based approach" (classification of systems by potential harm to rights/interests), "traceability" (audit logs and data provenance) and "data protection" (compliance with Argentina’s personal data protection framework).

Governance and Institutional Framework

Article II–III of the resolution assigns the Secretary of Institutional Coordination (Secretaría de Coordinación Institucional) responsibility for developing and following the Program, with operational cooperation from the Secretariats of Disciplinary Affairs and Technical & Administration and Human Resources. This design creates an internal governance triangle combining strategic direction, technical oversight and disciplinary review. The Program's remit is transversal: it must interact with prosecutors’ offices and technical teams across the MPF, channeling legal expertise and technical advice into project selection, evaluation and lifecycle management. The resolution situates the Program within existing institutional structures—rather than creating a separate independent regulator—and therefore relies on the PGN's authority to set institutional policy. For more on the MPF's public communications about the Program, see the MPF portal (Ministerio Público Fiscal) and the specific news release on the Program’s creation (MPF – UNESCO cooperation announcement).

Key Focus Areas

The Program’s mandated functions (Resolution clauses III a–i) identify several focus areas. First, institutional strategy and planning to integrate AI across MPF missions. Second, technical and ethical standard-setting for design, acquisition, implementation and evaluation. Third, needs assessment and prioritization of use cases, including administrative management, investigative analytics, pattern detection and citizen-facing services. Fourth, project intake and review: dependencies are invited to submit projects for analysis under criteria established by the Program. Fifth, creation of protocols and good-practice guides to operationalize human oversight, data governance, explainability, testing and documentation. Sixth, monitoring and evaluation: lifecycle supervision, outcome evaluation and impact assessment. Seventh, capacity-building: training programs for prosecutors, technical staff and administrators. Eighth, collaboration: the Program must cultivate ties with national actors (e.g., Agencia de Acceso a la Información Pública, Subsecretaría de Tecnologías) and international bodies (UNESCO, OECD), as well as academia, civil society and private-sector partners. These focus areas reflect an integrated governance posture: combining policy, technical standards, risk mitigation and transparency measures to ensure AI serves institutional objectives while protecting rights.

Implementation Framework

Operationalization depends on internal MPF procedures. The Program will develop criteria to evaluate project proposals, including risk classification (impact on rights and institutional interests), data quality and provenance, privacy compliance, security and resilience, human-in-the-loop design and auditability. Procurement and contracting of AI systems must integrate technical specifications, obligations for vendors to provide documentation, and contractual clauses addressing data protection and security. The resolution invites dependencies to present proposals and authorizes the Program to propose standardized protocols and templates for tendering, testing and deployment. Implementation will also require internal interoperability (e.g., with Coirón and SINOPE), secure data-sharing agreements, and alignment with national guidance from the Subsecretaría de Tecnologías de Información and the Agencia de Acceso a la Información Pública. Where systems implicate personal data, the Program must ensure compliance with Law No. 25.326 (Personal Data Protection) and internal MPF data policies.

Monitoring and Evaluation

The Program has explicit monitoring, follow-up and evaluation duties (III g). Monitoring includes technical performance checks, assessment of outcomes against objectives, documentation and audit logs for traceability, periodic reviews of model drift and updates, and evaluation of impacts on due process and rights. The Program will produce reports and recommendations to the PGN and may maintain a registry or inventory of approved AI projects and their key risk mitigations. Evaluation metrics should combine technical indicators (accuracy, false positives/negatives, robustness tests), process indicators (adherence to procurement and testing procedures) and rights-based metrics (disparate impact, privacy incidents, complaints received). The MPF also conducted an institutional survey (launched 11 August 2025) to inform baseline needs and priorities (MPF – institutional survey (11.08.2025)).

Penalties, Liability, and Appeals

The resolution itself does not prescribe new sanctions but locates oversight and disciplinary involvement within existing MPF frameworks; the Secretariats Disciplinaria and Technical will intervene according to their competencies. Consequently, legal and administrative liability, disciplinary sanctions and appeals will be applied under preexisting statutes and MPF internal rules (the resolution cites Law No. 27.148 and other resolutions as legal bases). Where projects contravene data protection obligations or cause rights violations, established administrative or disciplinary procedures may apply; civil or criminal liability would proceed according to applicable national legislation. The Program’s manuals and procurement templates are expected to embed contractual remedies and vendor obligations to allocate operational risk and support remediation.

Relationship to Other Instruments

The resolution cross-references a range of national and international guidance: UNESCO’s AI ethics recommendation and judicial AI toolkit; the OECD AI Recommendation; the EU AI legislative framework and Council of Europe instruments. Domestically, it builds on MPF prior resolutions (e.g., Coirón adoption: PGN 320/2017, mandatory electronic filing: PGN 22/2020) and national guidance on trustworthy AI issued by the Subsecretaría de Tecnologías de Información and the Agencia de Acceso a la Información Pública. The Program is explicitly designed to harmonize MPF practice with these instruments while remaining an internal coordination mechanism for the MPF's operational needs.

International Alignment

The Resolution acknowledges and seeks alignment with international normative developments: UNESCO (ethics recommendation, judicial toolkit), OECD (AI principles), EU Regulation (Regulation (EU) 2024/1689 and the Union’s AI law) and Council of Europe instruments on AI, rights and the rule of law. It also anticipates cooperation agreements (e.g., with UNESCO) to support training and ethical advisory services. International alignment aims to adopt principles such as human oversight, non-discrimination, privacy protection, transparency, technical robustness and accountability and to benefit from comparative testing frameworks and capacity-building programs.

Implementation Timeline

Sources and References