The OECD.AI Policy Navigator

Our policy navigator is a living repository from more than 80 jurisdictions and organisations. Use the filters to browse initiatives and find what you are looking for.

Use of artificial intelligence for the characterisation of the environmental regulatory burden of projects


Added by:   OECD analyst
Added on:   23 Sep 2026
Updated by:   OECD analyst
Updated on:   23 Sep 2026

The initiative quantifies the environmental regulatory burden in Chile by analysing environmental permits to identify the specific obligations generated. By providing objective evidence, it supports authorities and investors in creating a more transparent and efficient regulatory framework. The tool was developed to overcome the challenges of manual data processing, transforming thousands of pages of complex requirements into clear, comparable data for a more effective regulatory evaluation.

Name in original language

Uso de Inteligencia Artificial para la caracterización de la carga regulatoria ambiental de proyectos en desarrollo

Initiative overview

This initiative addresses the opacity and complexity of Chile's environmental regulatory system. Traditionally, the manual analysis of Environmental Qualification Resolutions (RCAs) has been prohibitively slow and costly due to the sheer length and lack of standardisation within these technical-legal documents. To tackle this, the initiative analysed a sample of 1,336 RCAs, representing all mining and energy projects authorised over the last decade. This analysis identified 220,000 individual enforceable obligations, providing a clear quantification and characterisation of the environmental burden imposed by Chilean regulations on major investment projects.

This application of GenAI transforms unstructured documents into auditable, comparable databases with unprecedented efficiency. It enables large-scale analyses that would have previously required an extensive team of experts working for several months.

Core objectives

  • Generate systematic evidence: Precisely dimension and characterise the regulatory burden on strategic sectors, mining and energy, by analysing over 220,000 obligations.
  • Explore emerging technologies: Evaluate the capabilities and limitations of GenAI in converting “opaque” regulatory information into actionable insights.
  • Promote transparency and efficiency: Advance towards a more streamlined regulatory system rooted in objective data and aligned with international standards.
  • Streamline enforcement: Identify and classify enforceable obligations to improve traceability and institutional compliance.

Evolution, implementation and scalability

The initiative is designed to evolve from an exploratory exercise into a permanent public policy tool:

  • Institutionalisation: The findings are expected to strengthen the legitimacy of public action, allowing the State to adapt continuously to a changing regulatory landscape.
  • Technological implementation: The results are already accessible via an interactive dashboard, offering users dynamic and detailed data visualisation.
  • Scalability: While this initial phase focused on mining and energy, the report establishes a replicable methodological framework applicable to any regulated productive sector.
  • Evidence-based reform: By identifying the 146 distinct laws that comprise the environmental inventory, this evidence serves as a foundation for future regulatory quality improvements and legislative discussions on system modernisation.

Other relevant details

Results, outcomes and impacts: The initiative processed 1,336 RCAs (2015–2024) in mining and energy, analysing 126,927 pages to identify 220,000 enforceable obligations and 146 environmental regulations. Its primary impact is transforming "opaque" documents into a structured, auditable database using Generative AI. This revealed a rising regulatory burden—increasing from 123 to 198 average obligations per project—and a shift toward "means-based" rather than "results-based" requirements. This GenAI model enables granular analysis by sector and phase, with future plans to scale it across all productive sectors as a permanent, data-driven public policy tool. Challenges: The use of AI faces risks of algorithmic bias, lack of model interpretability, and interpretational errors. A key methodological challenge was the variability and inconsistency in RCA structures—including randomised ordering, numbering errors, and irrelevant information—which hindered systematic extraction. The risks were managed through expert oversight to complement technical judgement. A rigorous validation protocol was implemented, utilising stratified random sampling and manual reviews by legal experts to ensure model accuracy. Lessons learned: A critical takeaway is that the quality of input data is essential for the model's success. Conditions for Success: Transparency, traceability, and replicability in design are indispensable. Success depends on technology acting as a support for, not a substitute of, the expert, as well as the availability of high-quality information repositories.

About the policy initiative


Category:

  • AI policy initiatives, programmes and projects

Initiative type:

  • AI use cases/projects in the public sector

Status:

  • Active

Start Year:

  • 2025

OECD AI Principles:


Other relevant urls: