Summary of the tenth session of the UN-REDD Programme’s Safeguards and Integrity Working Group in Latin America and the Caribbean
On 2 July, 2025, the tenth session of the UN-REDD Programme's Safeguards and Integrity Working Group in Latin America and the Caribbean was held under the title “Exploring Opportunities for Artificial Intelligence to Strengthen the Monitoring, Reporting, and Tracking of Safeguards”. This session focused on creating the space for dialogue on the emerging potential of Artificial Intelligence (AI) as a tool to support and enhance the technical and operational processes involved in the implementation of safeguards in REDD+ initiatives.
AI-related technologies are advancing rapidly and are already being applied in key sectors such as environmental and social risk management, climate governance, and the systematization and reporting of complex data. In this context, the Working Group convened an exploratory discussion aimed at sharing basic knowledge, identifying initiatives in development, reflecting on capacity-building needs in the region, and - above all - raising questions about what types of AI solutions could be useful for technical teams working on the design, monitoring, and reporting of REDD+ safeguards.
The growing interest in AI in safeguards contexts
The increasing demand for greater integrity in REDD+, along with the multiple sources and approaches to forest and climate finance - each with their own safeguards frameworks and requirements - can result in a growing workload for the safeguards teams and systems of countries and projects.
In this context, Artificial Intelligence (AI) is emerging as a tool with strong potential to support complex, repetitive or data-intensive tasks. Some of the most relevant functions highlighted during the session included the ability to automate the reading and analysis of lengthy technical documents, identify risk patterns based on existing data, generate preliminary versions of reports, and build assistance tools for technical field teams.
Beyond these specific functions, the discussion reflected a growing openness and interest in exploring AI’s potential, as well as a desire to better understand its technical requirements, limitations, and the conditions necessary for its ethical use, in line with principles of equity, participation and transparency.
The session made clear that AI is already part of the broader conversation on environmental governance and public management, and that there is a window of opportunity to collectively build a critical, strategic, and context-sensitive vision of its contribution to the strengthening of safeguards in the region.
Opportunities and challenges in applying AI to safeguards
Based on the participatory dialogue with representatives from different countries, several areas of interest and potential for incorporating artificial intelligence (AI) into safeguards work were identified:
Safeguards reporting: The use of AI was discussed as a means to support the assisted drafting of complex reports, such as Annual Progress Reports (APRs) for the Green Climate Fund or other large-scale projects, as well as Safeguards Information Summaries (RIS) submitted to the United Nations Framework Convention on Climate Change (UNFCCC). Mariano Cirone, from the UN-REDD UNEP team, also highlighted the potential of AI to facilitate the monitoring and preparation of safeguards reports for projects or public policies by automating the analysis of multiple standardized technical instruments, such as Environmental and Social Management Plans at the activity level. This approach allows for the preliminary handling of large volumes of information while promoting consistency and traceability.
Risk monitoring: Tools were discussed that can identify social and environmental risks based on reference documents or specific contexts. Aldo Franco, from the Paraguay + Verde Project, shared an ongoing initiative to train a specialized AI assistant capable of providing an initial automated review of risks and associated measures for the project. This tool aims to support a preliminary assessment based on standardized technical criteria, which must then be evaluated and complemented by the responsible technical team through validation processes, contextual analysis, and alignment with the applicable normative and operational frameworks.
Systematization and analysis of technical data: There was mention of the potential for AI to extract, classify, and organize key information from lengthy documents, making it easier to navigate content and generate inputs for decision-making.
Support for participatory processes and qualitative monitoring: Ideas also emerged regarding the use of AI to analyze narratives, open-ended comments, and community feedback in support of participatory monitoring systems.
Challenges to consider
Throughout the session, it was acknowledged that while interest in AI within the safeguards field is growing, significant challenges remain. Some of the key issues highlighted included:
- Strengthening specific technical capacities: Several participants emphasized the need to build the capacity of safeguards teams in the advanced use of AI, including the design of specialized agents, the development of effective prompts, and the critical evaluation of AI-generated results.
- Availability and accessibility of platforms: It was noted that while free versions of AI tools exist, more robust functionalities often require paid subscriptions or are tied to closed ecosystems, which can be a barrier in institutional settings with limited resources. Additional challenges related to connectivity and data storage were also identified, particularly in rural areas or those with limited technological infrastructure.
- Confidentiality and handling of sensitive information: Concerns were raised about the use of AI in contexts involving sensitive data, such as personal information, community statements, or internal documents not yet made public. The importance of having clear protocols on data privacy and governance was emphasized.
- Given the opportunities identified during the session, the Working Group recognized the importance of developing controlled testing environments where these technologies can be gradually evaluated in real-world contexts, with documentation of lessons learned and technical support. These pilot efforts are seen as key enabling conditions for the responsible and strategic integration of AI into national and subnational safeguards systems.
The environmental footprint of AI: A key dimension to consider
Beyond its benefits, the use of artificial intelligence carries significant environmental impacts. Training advanced models requires large volumes of energy, water, and critical materials, which can contribute to greenhouse gas emissions, water scarcity, and the generation of electronic waste. The extent of these impacts varies depending on the type of tool used and the technological environment in which it operates.
For this reason, it is essential to apply sustainability principles when integrating AI into safeguards systems. Possible strategies include opting for more efficient or open-source models, prioritizing platforms powered by renewable energy, and avoiding unnecessary applications.
For more information, see UNEP’s briefing note on AI and sustainability: https://www.unep.org/news-and-stories/story/ai-has-environmental-problem-heres-what-world-can-do-about
Voices from the Audience
José Carlos Fernández, Consultant, FAO
He highlighted the potential of AI to support the standardization of technical reports - a common challenge in processes involving multiple actors and institutions. He noted that AI tools could help produce documents with more consistent structure, language, and content, particularly in contexts marked by staff turnover or uneven technical capacities. He also mentioned technical challenges, such as converting documents into compatible formats, and stressed the importance of ensuring data privacy and traceability. He proposed setting “checkpoints” within the workflow to determine where AI could be used and where human validation is required. His contribution offered a critical and constructive perspective on the conditions needed for responsible institutional use.Jenny Carolina Grillo Gonzáles, UNEP Consultant for the Ministry of Environment of Colombia
She acknowledged the potential of AI-based interfaces to support safeguards systems but also raised questions regarding the cost and scalability of such tools at the national level. She asked how these solutions could be customized for government use without relying on costly developments, and suggested designing accessible, replicable models that build capacity within public technical teams.Teresita León, UNEP Team – Paraguay + Verde Project
She expressed enthusiasm for continuing to learn about the topic and for participating in future dialogue spaces. Her message reflected the genuine interest generated by the session - both for its technical insights and for the potential of AI to transform safeguards-related processes.
What is next? Key areas for collective progress
The use of AI in the field of safeguards is still in its early stages but initial experiences are beginning to emerge that explore its application as a sophisticated technical support tool. During the session, it became clear that both public and private actors are developing solutions aimed at automating tasks, systematizing information, analyzing risks, and generating reports more efficiently. These initiatives - though diverse in their approaches and stages of maturity - share common features: their experimental nature, use of accessible platforms, and the drive to test models adapted to the daily realities of safeguards work in REDD+ contexts.
In all cases, it was emphasized that these technologies do not replace professional judgment or participatory processes; rather, they can complement and enhance them. Participants also stressed the importance of moving forward through controlled pilot initiatives, with proper documentation of lessons learned, validation of results, and safeguards to ensure ethical and legal data management. These experiences represent an opportunity to develop technical solutions that are context-specific, effective, and aligned with the principles of safeguards.
The session concluded with a call to continue exploring, from a critical and collective perspective, how AI can be integrated in a useful, ethical, and strategic manner into safeguards systems across Latin America and the Caribbean.
Participants agreed that AI should not be seen as a one-size-fits-all solution or a technological imposition, but rather as a potential tool that must respond to the contexts, priorities, and regulatory frameworks of each country, in alignment with national climate and socio-environmental policies. In this regard, at least four key areas for further progress were identified:
- Strengthening technical capacities through applied learning
There is clear interest from many technical teams in better understanding how AI works, what tools exist, how to configure them, and how to evaluate their outputs. Designing hands-on training spaces, focused on real applications in safeguards processes, could be a key step in closing gaps and empowering teams. - Supporting pilot experiences and communities of practice
Ongoing cases show that individuals and teams are already experimenting with AI at different levels. Facilitating exchanges between these initiatives, documenting lessons learned, and generating collective feedback can accelerate the learning curve and help avoid common mistakes. Promoting regional collaborative spaces could be an effective way to sustain this exchange. - Fostering dialogue among diverse stakeholders
Integrating AI into environmental and social fields requires contributions from a wide range of perspectives - technical, legal, community-based, academic, and institutional. Creating spaces for horizontal dialogue among these actors will be essential to ensure that the design and use of these technologies are aligned with territorial realities.
Given the high level of interest expressed and the significant opportunities to strengthen safeguards teams' capacities, the UN-REDD Safeguards and Integrity Working Group will launch a Thematic Roundtable on Opportunities and Challenges of Artificial Intelligence Applied to REDD+ Safeguards Work starting in August 2025.
The Safeguards and Integrity Working Group of the UN-REDD Programme in Latin America and the Caribbean is a space made up of staff and experts working on the design, implementation, and monitoring of REDD+ safeguards in the region. Its goal is to share experiences, best practices, common challenges, and lessons learned to improve safeguards implementation, monitoring, and reporting—and to strengthen the environmental and social integrity of REDD+ in the region.
For more information about this work, please contact the UNEP team via Mariano Cirone at: [email protected]