This is the third and final post in our series sharing findings from our AI for Climate Action project. In our first post, we explored how organizations use a frugal approach to AI in technology design. In the second post, we examined the data governance decisions of organizations developing AI initiatives.
We close our blog series by exploring how AI for climate action is being imagined and shaped by social movements and environmental justice organizations.
Throughout the project, we explored what it means to incorporate an environmental lens into AI – from using these technologies to address environmental challenges to reducing their environmental costs. However, the cases in this research encouraged us to think about AI and climate action more broadly and through a more-than-human ecological perspective that recognizes the interdependence of people, knowledge, relationships, plants and animals within the ecosystems that sustain these technologies.
We found immense possibility in this reframing. It prompted us to consider how more responsible and ecosystem-aware approaches to AI could themselves be a form of climate action, including in initiatives that are not explicitly designed to address environmental challenges. It raised questions about what sustainability means in AI systems and how it can encompass the long-term flourishing of communities and healthy ecosystems. Overall, the approach helped us map the close connections between environmental protection and caring for the people, knowledge and relationships tied to the territories we seek to protect.
In this final blog post, we explore how AI systems support and advance climate action, how organizations center care in decisions, and the role of social movements in shaping more sustainable AI. These insights can offer lessons for organizations whose work may not be explicitly framed as climate action, but who are keen to align technological choices with more just and sustainable futures.
The diverse ecosystem of AI and climate action
The project highlights the ways AI and climate action intersect. Much of the conversation at this intersection focuses on how AI can support environmental data monitoring, prediction and analysis. For example, Flood Guard in Uganda used flood forecasting to share warnings and make predictions to support waterborne disease prevention in the Kampala metropolitan region. Likewise, Situational Intelligence Open Source Software (Siti OSS) in Indonesia uses its platform to facilitate a two-way exchange of information.
However, our research also brought attention to initiatives where protecting community knowledge, cultural practices and relationships was closely connected to environmental stewardship. Examples include, Saving Voices, a project developing AI voice models to preserve Indigenous languages, and an AI of Our Own, an initiative supporting cultural preservation across Cambodia, Ghana and South Africa. Both projects recognized the connections between protecting cultural knowledge and the environments in which communities build and share it. For Saving Voices, as community members saw value in the preservation of their local language, they began considering what else it could make possible, such as the preservation of folk stories, biodiversity knowledge and observations about the local environment.
Other projects use AI tools to support learning and advocacy, creating opportunities for communities to engage more actively in environmental protection. In Lebanon, projects like Taxonomy Meets Tech or in Mexico, SKILIKET, used citizen science approaches to involve community members in conservation, while exploring how the process could support the understanding of the environment. In Ghana, the Digital Literacy and AI for African Crop Disease Detection project shared information through its pest detection app to encourage organic and environmentally friendly responses to crop diseases and lower reliance on pesticides and insecticides. Similarly, Farmers for Forests in India supports farmers in strengthening their resilience to climate change by planting trees alongside crops, while using monitoring technologies to help them participate in and benefit from carbon markets.
Finally, for other organizations, AI adoption extended to broader struggles for land rights, territorial defence and environmental justice. For example, in Mexico, Diversa Studio co-designed a tool with the Yaqui community to address questions of water governance. In Kenya, the Voice of the Ogiek Messenger, used a combination of SMS, AI and low-tech tools to share information on land rights. These initiatives are shaped by existing struggles for environmental justice, which inform the purposes and priorities of the technologies being developed towards protecting communities from deforestation and displacement.
Looking at these cases together, the value of AI was measured by the possibilities it created for the people and communities using it. Many of these projects, as we explored in our first post, also employed frugal approaches, such as offline capabilities, edge hardware or smaller models – sometimes this was an intentional decision to reduce environmental impacts, while others as response to local infrastructure and resource constraints. These experiences encouraged us to consider how sustainability can be reflected in the choices organizations make throughout the lifecycle of an AI initiative, including decisions about how technologies are developed, maintained, expanded or eventually retired.
Growing and retiring AI systems with care
Thinking about sustainability from a broader ecological perspective also invites us to reconsider what it means for an AI initiative to grow and remain useful over time. While scaling up is often understood as expanding the reach and impact of a tool, the organizations in our research encouraged us to think about growth in different ways. For some, growing with care meant making technologies and resources accessible for others to adapt, strengthening relationships with other organizations and social movements, and recognizing when a tool had served its purpose and was no longer needed.
For some organizations, these decisions began with selecting technical architectures that could facilitate uptake by other groups and create a fertile ground for collaboration and solidarity. In Libya, Mozn, a youth-led weather-monitoring initiative developed by Roaya for Astronomy and Space Applications Foundation, prioritized open-source approaches and made weather data from nearly 55 monitoring points freely available:
“I kept asking: with scarce resources, can this be rebuilt? If another country installed weather stations and collected its own data, could it put that data into the system?”
This approach allows researchers and organizations facing similar environmental challenges to build on existing resources and adapt them to their own contexts, including those with limited resources, without relying on increasingly sophisticated infrastructure. Similarly, in India, Saving Voices, uses open-source, frugal AI models by design, allowing other researchers to build on its work through collaboration and co-creation:
“This is not only for Indigenous communities in India. Once the interface and technology are developed, they can be adapted for Indigenous communities anywhere in the world. The tool will be open source so that other communities can co-create with it. For example, an Indigenous community in Africa and an Indigenous community in India could use the technology to create something together. It could be music, cultural knowledge or practices such as seed preservation: How are seeds preserved in one community, and how are they preserved in another? This kind of knowledge exchange and co-creation could take place.”
Sustaining these initiatives also involves ensuring that communities have the knowledge and opportunities to make informed decisions about their use. Organizations such as An AI of Our Own and PetaBencana.id have developed educational materials, workshops and spaces for ongoing conversations with participating communities to explore the possibilities and limitations of AI and strengthen people’s capacity to engage with technologies responsibly. Building these capacities can also facilitate peer learning and enable communities to decide whether to use or retire a technology on their own terms.
Conversations about growth also opened up reflections on degrowth and the lifecycle of AI initiatives, raising questions about whether maintaining a technology indefinitely is always desirable. For example, the team behind Voice of the Ogiek Messenger considered the possibility of retiring the tool from the outset. For the team, the success of the initiative was tied to whether it has supported the community’s wellbeing and whether people can continue to flourish after the technology is no longer needed.
“I am aware that tools should exist for the time they need to exist and serve the purpose they need to serve. So it might not serve them for too long, but as long as it was there, it did what it did and it did not take more from the community than it gave…. We see flourishing, that the community flourished and they are still flourishing after the tool stopped working. I think that’s the idea. We build tools so they can benefit communities and when they don’t need them, then we are successful.”
These experiences invite us to understand the success and sustainability of AI initiatives, and how we can nurture the knowledge and capacities of communities and social movements so they can continue to shape the technologies intended to support them over time.
Imagining just AI futures led by social movements
These reflections on sustainability also invite us to consider whose visions of climate justice are shaping AI technologies. Social and environmental justice movements have long questioned the environmental costs and extractive dynamics of technological development. Through our research, we encountered initiatives where movement members were also becoming directly involved in imagining and building technologies grounded in their political priorities.
In Brazil, IARAA was conceived, designed and built by activists and movement members from the Landless Workers’ Movement (MST) and the World March of Women, with support from Baobá, the Association for International People’s Cooperation. Their involvement created opportunities to develop technical skills within the movements and explore how digital technologies could reflect their political commitments and environmental concerns. It also connected their struggles for environmental wellbeing with broader questions of digital sovereignty and the ability of movements to shape the technologies they use.
“We didn’t just develop IARAA; we began to view this issue as a whole, as a strategic issue. And because of that, we also started conducting training sessions with activists from organizations here in Brazil, where we go over some of the training journey we went through ourselves that involved learning not only political critique but also how to master the technical aspects.”
According to the team, this experience brought a new dimension to how movement members engaged with AI. Their earlier discussions had focused on the environmental costs of these technologies and their extractive dynamics, however through their involvement, they were able to explore how the organizing practices of social movements could inform more responsible and just technological development. This created opportunities for movement building and helped other activists see themselves as capable of shaping and developing AI technologies.
The team behind Voice of the Ogiek Messenger shared a similar vision of technological futures shaped by communities, as they hope their approach can encourage others to rethink AI development:
“My imagination is that many people would embrace it. Not just the tool, not just what we build, but if they can take those ideas and rethink them and build different tools for different contexts. As long as we can have many more people building low power AI tools, I think that’s how I see the future. So my future is if people can rethink AI, if more people can go back and say, hey, how can we make AI less costly, less extractive, causing less harm to communities?”
This vision puts the knowledge and priorities of social movements, organized collectives and communities at the center of these conversations. While AI systems can be allies in the work of build more just and sustainable futures, it is people and their collective action who will drive the changes needed to make these futures possible.
We would like to give special thanks to the twelve organizations who generously shared their stories for this study, as well as our regional researchers: Madhuri Karak, Radhika Jhalani, and Yosr Jouini, who conducted the desk research and interviews in Asia and North Africa.