Chile - Algorithmic Transparency Recommendations (372/2024)

Overview

Resolución Exenta N° 372 (dated 12 August 2024 and published in the Diario Oficial on 30 August 2024) approves the text of the "Recomendaciones del Consejo para la Transparencia sobre Transparencia Algorítmica". The measure is an advisory, non‑binding set of recommendations targeted at public bodies subject to the Law on Access to Public Information (Law N°20.285) and related rules. It was adopted by the CPLT to respond to the expanding use of systems of automated and semi‑automated decisions (SDA) across public agencies and to align transparency practice with the protection of fundamental rights. The full text appears in the official publication for the Diario Oficial and is distributed by the CPLT; see the official resolution document at CPLT / Diario Oficial PDF (Resolución Exenta N° 372) and the CPLT announcement at CPLT news release.

Definitions

The recommendations define and rely on key concepts: "systems of decisions automated and semi‑automated (SDA)" to include algorithmic systems, models and decision support systems used to make or inform public decisions; "transparency proactiva" (proactive transparency) to indicate systematic public disclosure through transparency portals; "black‑box" systems to describe models whose internal logic is not directly interpretable by non‑developers; and "sujetos obligados" as those entities obliged under Law N°20.285 to provide public information. The text situates these terms within existing Chilean laws (including Law N°20.285, Ley N°19.628 on privacy, and Ley N°21.180 on digital transformation) and international standards cited in the preamble.

Governance and Institutional Framework

The CPLT positions itself as a promoter of transparency best practices and issues these recommendations pursuant to its authority under Article 33(e) of Law N°20.285. The resolution refers to collaboration with other State instruments and prior guidance (for example, the Ministerial Oficio Circular N°711, 11 December 2023) and situates the recommendations as complementary to national initiatives (including the National AI Policy). The CPLT encourages internal governance measures within agencies—appointment of responsible officers for SDA transparency, maintenance of inventories, version control and audit trails—and coordination between data protection, cybersecurity and procurement units. For reference and formal publication consult the official resolution at the Diario Oficial PDF: Diario Oficial (CVE 2537198) and the CPLT guidance repository at CPLT guide page.

Key Focus Areas

The recommendations are structured around several substantive focus areas. First, proactive transparency: agencies should publish a standardised set of information about any SDA that affects public services, benefits, eligibility or rights—this includes the name of the system, objectives, areas of application (services/trámites), general architecture or model type, decision logic description, categories and sources of data used, known impacts and risk mitigation measures, procurement and contractor information, update cadence and contact points for human review. Second, accessibility and communication: information must be concise, intelligible and available in multiple formats to reach diverse audiences. Third, auditability and documentation: agencies should retain technical documentation, model specifications, training/validation data summaries and logs to support ex post audits and oversight. Fourth, balance with confidentiality: when confidentiality, privacy or IP limit disclosure, agencies should document legal bases for non‑disclosure and publish as much information as possible. Fifth, handling opaque or proprietary models: preference is given to systems enabling monitoring and explainability; when black‑box solutions are used, agencies must document compensating controls (independent audits, human oversight, impact assessments). Sixth, rights protection: guidance emphasises non‑discrimination, due process and privacy safeguards and aligns with pending personal data protection reforms.

Implementation Framework

The resolution recommends a practical, staged approach for adoption. Agencies are advised to create internal inventories and to integrate a "Transparencia Algorítmica" section into their active transparency portals, with periodic updates (the CPLT suggests monthly reporting for active systems). The CPLT’s later technical guide (2025) provides templates and examples, recommended metadata fields and illustrative entries (e.g., the School Admission System — MINEDUC). The implementation framework favours proportionality: agencies with greater technological maturity are encouraged to publish more granular technical documentation, while smaller agencies should at minimum publish plain‑language descriptions and contact mechanisms. Recommended operational steps include: appointing a responsible transparency officer; conducting an initial mapping; applying a minimal metadata schema (registry entry, function, scope, data categories, contact information); establishing update and audit processes; and documenting decisions on non‑disclosure. See the CPLT adoption guide at CPLT guide for implementation tools and templates.

Monitoring and Evaluation

Monitoring is framed as both the CPLT’s advisory follow‑up role and the responsibility of each agency to self‑report. The CPLT recommends that agencies publish version histories, update logs and impact monitoring (including identified errors, bias incidents and corrective measures). The CPLT envisages periodic reviews and encourages pilot testing, independent audits and cross‑agency knowledge sharing. While the recommendations are not coercive, failure to adopt transparent practices may be flagged in CPLT oversight reports and could inform subsequent formal interventions under Law N°20.285. The CPLT and other oversight bodies are encouraged to use published registries to analyze sectoral adoption rates and identify systemic risks.

Penalties, Liability, and Appeals

These recommendations do not by themselves create new criminal or administrative sanctions. The CPLT explicitly issues them as non‑binding guidance that informs how agencies should meet transparency obligations under existing laws. Nevertheless, transparency failures can be the subject of requests, complaints and judicial review under Law N°20.285 and related statutes; agencies may be required to justify non‑disclosure based on specific legal grounds. The document advises agencies to document decisions that restrict disclosure to facilitate external review. Potential consequences for systemic non‑compliance include reputational damage, CPLT findings, and referral to competent authorities when other legal infractions (e.g., data protection breaches) are identified.

Relationship to Other Instruments

The resolution cross‑references and complements multiple Chilean instruments: Law N°20.285 (Access to Public Information), Ley N°19.628 (Protection of Private Life), Ley N°21.180 (Digital Transformation of the State), Oficio Circular N°711 (Ministries guidance on AI use, Dec 2023), and emerging data protection and cybersecurity frameworks (including Ley N°21.663 on cybersecurity references). The recommendations explicitly cite the 2021 study with Universidad Adolfo Ibáñez as background research (identifying 92 SDA across 42 agencies) and position the guidance alongside the National AI Policy and ongoing legislative work on data protection and AI regulation. By design, the recommendations are interoperable with procurement rules and sectoral guidance (finance, health, social protection) and encourage harmonization through publication standards to be adopted in state transparency portals.

International Alignment

The CPLT anchors the recommendations in international norms and comparative developments, citing instruments such as UNESCO’s Recommendation on the Ethics of Artificial Intelligence and relevant practices from data protection and access‑to‑information authorities internationally. The text emphasizes alignment with international standards on explainability, non‑discrimination and human oversight, and encourages participatory development and knowledge exchange with peer regulators. The CPLT’s international engagement and publications related to these recommendations are described on their site and in conference reports; see CPLT materials at Consejo para la Transparencia.

Implementation Timeline

Sources and References