Data Governance · 4 min read
Chile Lacks Sector-Specific Rules for Healthcare AI, Review Finds
A Frontiers in Digital Health review finds Chile's pre-AI statutes cannot catch proxy-variable discrimination across its FONASA-ISAPRE divide, and proposes seven governance dimensions.
Key takeaways
- FONASA covers 76.5-78% of Chile's population; ISAPRE private insurers serve wealthier, younger enrollees.
- Law No. 21.719's automated-decision rights (human intervention, explanation, review) become enforceable December 1, 2026.
- Chile's AI Bill (Bulletins 15.869-19 and 16.821-19) remained under legislative consideration as of October 2026.
- The review proposes a seven-dimension governance framework covering bias assessment, auditing and human oversight.
- EU deferred high-risk AI compliance to December 2027 (Annex III) and August 2028 (Annex I) via Regulation (EU) 2026/1744.

Chile has no coherent, sector-specific governance framework for clinical artificial intelligence, despite a constitutional right to health and a stack of statutes written before automated systems entered hospitals. That is the central finding of a qualitative socio-legal review by Bárcenas Vidal, Martínez Montenegro and Fuertes Iglesias, published in 2026 in Frontiers in Digital Health (DOI: 10.3389/fdgth.2026.1807073).
The problem is coordination, not absence. Chilean law protects equality, non-discrimination, personal data and patient rights. But those safeguards sit fragmented across general administrative domains. Nothing articulates obligations across the development, deployment and auditing lifecycle of clinical AI.
Why does structure matter for algorithms?
Chile runs a dual healthcare system. The public National Health Fund, FONASA, covers roughly 76.5% to 78% of the population, disproportionately people with lower incomes and higher health needs. The private insurers, ISAPRE, serve wealthier and younger demographics through individual contracts that incentivize risk selection and cream-skimming (enrolling profitable patients while excluding costly ones). The result: persistent asymmetries in access, funding and quality by socioeconomic status.
Models trained on data from this divided environment will encode its distortions. The authors warn that deploying algorithmic systems without governance interventions risks reinforcing historical disparities in patient care.
What is the difference between bias and discrimination?
The review draws a distinction data teams should internalize. Algorithmic bias means systematic distortions from training-data deficiencies, design parameters, optimization functions or unrepresentative sampling. Algorithmic discrimination means those distortions produce unequal treatment of individuals or protected groups — and it requires no discriminatory intent. The review cites the classic case: historical healthcare expenditure used as a proxy for clinical need. Disadvantaged populations spend less on care because of access barriers, not lower illness. Systems relying on financial proxies systematically underestimate their urgency.
Which laws apply — and where do they fall short?
- Article 19 No. 2 of the Constitution guarantees equality before the law; Article 19 No. 9 establishes the right to health protection and state oversight of equal access to recovery and rehabilitation.
- Law No. 20.609 (the Zamudio or Anti-Discrimination Law) provides judicial recourse against arbitrary discrimination.
- Law No. 20.584 codifies patient rights to dignified care, clinical information and non-discriminatory treatment.
These statutes target individual, identifiable acts and overt intent. They lack evidentiary mechanisms for indirect discrimination, automated opacity, or population-level disparate impact. Baseline data governance under Law No. 19.628 does not address automated risk profiling or algorithmic decision-making across the software lifecycle.
What is coming next?
Chile's legislature is moving. Law No. 21.719 (2024) creates a Personal Data Protection Agency and Article 8 bis rights for automated decisions — human intervention, transparency, explanation and review — enforceable December 1, 2026. A consolidated AI Bill (Bulletins 15.869-19 and 16.821-19), still under consideration as of October 2026, mirrors the EU AI Act's risk-tiered approach and adds equity, non-discrimination and cultural diversity principles under Article 4. None of this is yet enforceable health-sector law.
The institutional backdrop is turbulent. Supreme Court rulings in November 2022 forced private insurers to apply a standardized risk-factor table and reimburse excessive charges. Law No. 21.674 (May 2024), the Short Isapres Law, implemented compliance and created a Complementary Coverage Modality within FONASA. Bill No. 17.375-11 (January 2025) aims to modernize public services. The Superintendence of Health confirms Banmédica, Colmena, Consalud, Nueva Masvida and Vida Tres still operate under heavy financial adjustment.
What should teams do?
The authors propose a seven-dimension framework: transparency and explainability, accountability, human oversight and contestability, bias and equity assessment, data governance, auditing with continuous monitoring, and institutional literacy — all deployable within existing legal remits.
One caution travels beyond Chile. The EU itself deferred high-risk compliance deadlines — Annex III to December 2027 and Annex I to August 2028 under Regulation (EU) 2026/1744 — to accommodate institutional readiness. Effective oversight, the authors argue, depends on post-market surveillance and technical guidance, not formalistic declarations, and Chile cannot simply transplant European mechanisms given domestic resource disparities.
The review searched Scopus, Web of Science, PubMed, SciELO, HeinOnline and Google Scholar plus institutional repositories between January and March 2026. It tested no tools, validated no models and measured no discrimination prevalence. Its authors concede the limits: no non-public documents, no stakeholder interviews, no clinical testing. Future work, they conclude, must evaluate algorithmic governance empirically to preserve patient rights and substantive health equity.
doi.org (Original)
More from Rebecca Stone
Show full bio
Senior reporter covering regulation, policy and their impact on Analytics Digest is a professional daily at Analytics Digest.
12 articles