OECD Principles
Referential
- Type
- Reference
- Scope
- Country
- Status
- In force
- Published
- What it contributes
- Ethical and political framework for trustworthy AI.
- What it does not establish or solve
- Provides no institutional procedures.
Country framework
Costa Rica AI country framework
Architecture of public policy, regulation, technical guidelines and institutional adoption of AI within the Costa Rican State.
Costa Rica already has principles, strategy, technical guidelines and institutional adoption. The pending gap is turning that framework into shared, verifiable and actionable procedures for public institutions.
Last update: October 2026
Quick indicators
What the country framework already shows, in verifiable numbers.
- Record
- 1
- National strategy
- ENIA 2024-2027
- Record
- 1
- Action Plan
- Inter-institutional implementation
- Record
- 1
- AI chapter in the CNTD
- Mandatory technical guidelines
- Record
- 7
- Legislative files
- None approved yet
- Record
- 16
- Institutions with documented initiatives
- Institutional catalog across three evidence layers
- Record
- 39
- Initiatives mapped
- 9 verified, 14 under review and 16 ecosystem records
Country framework architecture
The AI framework in Costa Rica does not rely on a single document. It is built from layers with different function, scope and institutional force. Some guide, some bind, some are in process and others show real adoption.
- Referential
Layer 1
International principles
Instruments
OECD, global recommendations and public-policy tools.
Function
Frame the national conversation on trustworthy AI, rights, transparency, robustness, accountability and responsible innovation.
Scope
Country level
Force
Referential
Gap it leaves
Does not translate principles into concrete institutional procedures.
- Guiding
Layer 2
National strategy
Instruments
ENIA 2024-2027 and its Action Plan.
Function
Defines priorities, country vision, action lines and inter-institutional commitments for AI development and adoption.
Scope
National and inter-institutional
Force
Guiding and coordinating
Gap it leaves
Does not by itself set the daily procedure each institution must follow when adopting, procuring, evaluating or scaling AI systems.
AILA 2026, published by UNDP and MICITT, assesses national readiness. It informs this reading but is neither an implemented target nor a binding guideline.
- Mandatory
Layer 3
CNTD, AI chapter
Instruments
National Digital Technologies Code, Executive Decree 44507-MICITT, Chapter 7 on Artificial Intelligence.
Function
Sets technical and governance criteria for public-sector digital initiatives, including risk management, security, maintenance, technical standards, sustainability, monitoring and incident response.
Scope
Public sector
Force
Mandatory for State digital initiatives and projects
Gap it leaves
Sets guidelines and controls, but does not yet function as a single operational playbook with standardized forms, roles, evidence, risk thresholds and approval flows.
- Not in force
Layer 4
Bills in process
Instruments
7 legislative files: 4 whose primary subject is AI and 3 directly related to digital advertising, deepfakes, personal traits, digital creations and coordinated manipulation.
Function
Aims to establish rights, duties, institutional limits, responsibilities and control mechanisms.
Scope
National
Force
Not yet in force
Gap it leaves
With no approved law, Costa Rica operates without a specific AI legal framework. Even with a law, an operational layer translating legal duties into institutional procedures will still be necessary.
- Operational
Layer 5
Institutional initiatives
Instruments
39 AI-related initiatives across 16 institutions: 9 verified adoptions, 14 under review and 16 ecosystem and capacity records.
Function
Documents AI-related systems, pilots, plans and capabilities in the public sector. The current inventory includes 39 initiatives of different types and maturity; only 9 meet the verified-adoption rule.
Scope
Institutional and sectoral
Force
Varies by initiative
Gap it leaves
Initiatives have different levels of evidence, standards and governance, without a common public methodology across the State.
In addition to production projects, there are sectoral governance instruments such as the Basic Guidelines for the use of Authorized Generative AI in the Judiciary. These are relevant but do not amount to a national framework applicable to the entire State.
- Pending
Layer 6
Common procedure pending
Instruments
National public-AI playbook (pending).
Function
Turn principles, strategy and guidelines into verifiable procedures for institutional teams.
Scope
Costa Rican public sector
Force
Shared operational guidance, ideally linked to the CNTD and the digital-government model
Gap it leaves
Risk, procurement, human oversight, traceability, audit, pilot-to-production transition, incidents, user rights and service continuity.
Costa Rica already has a framework. The next step is turning it into shared operational capacity.
AI governance milestones in Costa Rica
Unlike the institutional adoption timeline, this view shows the evolution of the country framework: documents, guidelines, regulation, international engagement and strategic decisions.
OECD AI Principles: first intergovernmental standard for trustworthy AI, with Costa Rica among the adherent countries in Latin America
First legislative files on AI submitted to the Legislative Assembly
Publication of the National Artificial Intelligence Strategy (ENIA 2024-2027)
Incorporation of Chapter 7 on AI into the CNTD (Decree 44507-MICITT)
ENIA Action Plan
- 2025
Progress of substitute texts and legislative reports on AI
UNDP and MICITT publish AILA 2026, a national readiness assessment across three pillars
- 2026
Consolidation of the public inventory of AI projects across the State
- Pending
General AI law or regulatory framework approved
- Pending
National operational playbook for public AI adoption
What each document or instrument contributes
An assessment does not carry the same force as a strategy, technical guideline or law. This matrix distinguishes what exists, who it applies to, what it contributes and what it does not establish.
ENIA 2024-2027
Guiding
- Type
- Strategy
- Scope
- Country
- Status
- In force
- Published
- What it contributes
- Country priorities and vision for AI.
- What it does not establish or solve
- Does not define a daily operational protocol.
ENIA Action Plan
Guiding
- Type
- Strategic implementation
- Scope
- Inter-institutional
- Status
- In force
- Published
- What it contributes
- Inter-institutional commitments and tasks.
- What it does not establish or solve
- Does not standardize daily execution per institution.
AILA 2026
Referential
- Type
- Assessment
- Scope
- Country
- Status
- Published
- Published
- What it contributes
- Assesses national readiness across the ecosystem, AI for government, and regulation and ethics.
- What it does not establish or solve
- Does not verify individual project implementation or create institutional obligations.
CNTD, Chapter 7 AI
Mandatory
- Type
- Technical guideline
- Scope
- Public sector
- Status
- In force
- Published
- What it contributes
- Technical criteria and governance controls.
- What it does not establish or solve
- Does not descend to a detailed institutional playbook.
AI bills
Not in force
- Type
- Regulation
- Scope
- Country
- Status
- In process
- Published
- What it contributes
- Eventual legal framework of rights and duties.
- What it does not establish or solve
- Does not solve current public operations.
Basic Guidelines for the use of Authorized Generative AI in the Judiciary
Operational
- Type
- Sectoral
- Scope
- Judiciary
- Status
- In force
- Published
- What it contributes
- Authorized internal use of generative AI within the Judiciary.
- What it does not establish or solve
- Does not apply to the State at large.
Operational governance gaps
Costa Rica already has a base of public policy and technical guidelines. The next challenge is moving from framework documents to installed institutional capacity.
- There is no national AI playbook for the public sector.
- There is no common format for AI-system inventory.
- There is no shared methodology for classifying use-case risk.
- There are no common public-procurement criteria for AI solutions.
- There is no national human-oversight and traceability protocol.
- There is no public AI-incident reporting mechanism.
- There is no institutional compliance dashboard.
From policy to documented implementation
The country framework shows rules, strategies and guidelines. The institutional inventory brings together systems, pilots, plans and capabilities with different levels of evidence. One explains the governance architecture; the other supports implementation tracking.
Sources and inclusion criteria
This page includes national documents, international instruments, technical guidelines, legislative files and documented institutional initiatives. Announcements or pilots without evidence of execution should be shown as such, not as verified adoption.
Sources
- Official MICITT documents
- Executive Decree 44507-MICITT and CNTD
- Legislative Assembly
- OECD
- ILIA
- Public institutions
- AI Observatory Costa Rica own mapping
Inclusion criteria
- Verifiable public document as primary evidence.
- Inclusion proportional to instrument scope (country, public sector, institution).
- Explicit distinction between referential, guiding and mandatory force.
- Explicit distinction among operational systems, pilots, plans and capabilities; an official mention does not equal execution.