Costa Rica - AI Regulation Bill

Overview

The Artificial Intelligence Regulation Bill was introduced in Costa Rica's Legislative Assembly on May 30, 2023, by Congresswoman Vanessa Castro and colleagues. The bill gained international attention for its unprecedented approach: the entire text was drafted with assistance from ChatGPT-4, making it one of the first AI-generated legislative proposals in the world. The bill proposes establishing a comprehensive regulatory framework for artificial intelligence in Costa Rica, including the creation of a supervisory authority and requirements for accountability, explainability, and bias prevention. The initiative reflects Costa Rica's proactive approach to AI governance and positions the country among Latin American leaders in developing AI regulatory frameworks.

Definitions

The bill provides foundational definitions for AI regulation. Artificial Intelligence Systems refers to computer systems capable of performing tasks that typically require human intelligence, including learning, reasoning, and decision-making. The Competent Authority is the proposed regulatory body responsible for overseeing AI systems. Accountability refers to the requirement that organizations deploying AI systems take responsibility for their outcomes. Explainability means AI systems must be capable of providing understandable explanations of their decision-making processes. Bias refers to systematic errors that produce unfair outcomes, particularly discrimination against protected groups. High-Risk AI Systems are those with significant potential impacts on fundamental rights or safety.

Governance and Institutional Framework

The bill proposes establishing a new Competent Authority specifically responsible for AI oversight in Costa Rica. This regulatory body would have independence to carry out its supervisory functions without political interference. The authority would be empowered to develop implementing regulations, issue guidance on AI compliance, receive and investigate complaints about AI systems, conduct audits of AI deployments, and impose sanctions for violations. The governance structure anticipates coordination with existing regulatory bodies including data protection authorities, consumer protection agencies, and sector-specific regulators. The Legislative Assembly would maintain oversight of the authority through reporting requirements and budget approval.

Key Focus Areas

• Establishment of Supervisory Authority: Creating an independent regulatory body with powers to oversee, audit, and enforce AI compliance.
• Accountability Requirements: Mandating clear responsibility chains for AI system outcomes and decisions.
• Explainability Standards: Requiring AI systems to provide understandable explanations of their decision-making processes.
• Bias Prevention: Implementing measures to identify, assess, and mitigate algorithmic bias and discrimination.
• Human Rights Protection: Ensuring AI development and deployment respects fundamental rights and dignities.
• Supervision and Auditing: Establishing frameworks for regulatory oversight and compliance verification.
• Complaint Mechanisms: Creating accessible processes for citizens to report AI-related harms or concerns.
• Transparency Requirements: Mandating disclosure about AI use and capabilities to affected individuals.
• Risk Assessment: Requiring evaluation of AI risks before deployment in high-stakes contexts.
• Enforcement Powers: Granting the competent authority ability to impose sanctions for violations.

Implementation Framework

If enacted, the bill would require establishment of the competent authority within a specified timeframe following passage. The new regulatory body would then develop detailed implementing regulations and technical standards for AI compliance. Organizations deploying AI systems would have transition periods to achieve compliance with new requirements. The implementation framework anticipates phased rollout with initial focus on high-risk AI applications before expanding to broader AI governance. Capacity building within government, support for regulated entities, and public awareness campaigns would accompany regulatory implementation.

Monitoring and Evaluation

The bill establishes monitoring mechanisms through the competent authority's supervisory powers. Article 6 specifically addresses Supervision and Auditing, empowering the authority to examine AI systems for compliance with regulatory requirements. Regular reporting by AI deployers would provide ongoing visibility into AI use across sectors. The authority would maintain registers of AI systems subject to oversight. Periodic assessments would evaluate the effectiveness of the regulatory framework and identify needed adjustments. International benchmarking would compare Costa Rica's approach with evolving global best practices.

Penalties, Liability, and Appeals

The bill contemplates sanctions for non-compliance with AI regulatory requirements, though specific penalty amounts would likely be detailed in implementing regulations. Violations related to bias, lack of transparency, or failures of accountability could result in administrative fines, orders to cease AI operations, or requirements to implement corrective measures. Criminal liability might apply in cases of intentional misconduct causing significant harm. Organizations would be liable for harms caused by their AI systems, with potential for joint liability between developers and deployers. Appeals processes would allow regulated entities to challenge regulatory decisions through administrative and judicial review mechanisms.

Relationship to Other Instruments

The bill would operate alongside Costa Rica's existing legal framework including constitutional rights protections, data protection legislation, consumer protection laws, and sector-specific regulations. It complements the National Artificial Intelligence Strategy (ENIA) 2024-2027 which provides policy direction for responsible AI development. The legislation would interact with Costa Rica's digital transformation initiatives and e-government programs. International commitments under human rights conventions inform the bill's rights-protective provisions. The bill represents one of several AI-related legislative proposals under consideration in the Legislative Assembly.

International Alignment

The bill reflects influence from international AI governance frameworks, particularly the approach taken by the EU AI Act with its emphasis on risk-based regulation and transparency requirements. The principles of accountability and explainability align with the OECD AI Principles. Human rights provisions reflect commitments under the American Convention on Human Rights and other international instruments. The bill positions Costa Rica to participate in regional harmonization efforts for AI governance in Latin America. The innovative use of AI in drafting the legislation itself demonstrates Costa Rica's engagement with emerging technology while raising novel questions about AI governance.

Implementation Timeline

Sources and References