By Stuart Kerr, Technology Correspondent, LiveAIWire
AI in climate justice debates has moved from theoretical concern to documented pattern, and Google DeepMind’s AlphaEarth Foundations, launched in July 2025 as a virtual satellite capable of mapping the planet’s entire terrestrial surface at 10-metre resolution, sits squarely at the centre of it. AlphaEarth is a genuine scientific achievement, integrating satellite imagery, radar, and climate data into a single digital representation researchers can query on demand. The unresolved question is who controls that representation, and whether Indigenous communities whose land it maps get any say in how the resulting data gets used.
That question is not hypothetical. A 2023 research paper on Indigenous data sovereignty, applying the CARE principles, Collective Benefit, Authority to Control, Responsibility, and Ethics, documents a recurring pattern: well-intentioned data projects that map, monitor, or analyse Indigenous land without meaningful consultation end up causing real harm, even when the stated goal is conservation or humanitarian benefit. One case study in the paper describes a European NGO’s water-monitoring project in Burundi that publicly shared geographic data about a community without adequately understanding local concerns, violating collective privacy and eroding the trust the project was meant to build.
Why Neutral-Looking Data Isn’t Actually Neutral
The core argument researchers make is that describing satellite-derived land data as neutral obscures a real decision embedded in how that data gets framed. A parcel of forest categorised by an AI system as under-utilised, low-yield, or available for development carries an implicit judgement about value, one that reflects whichever priorities were built into the model rather than the relationship an Indigenous community actually has with that land. The CARE principles framework argues explicitly against treating Indigenous data reuse the same way scientific data reuse works under the more common FAIR principles, since FAIR focuses purely on making data findable and reusable, without addressing who benefits from that reuse or whether the people the data describes had any say in how it gets used.
What This Means for How Conservation Projects Should Actually Work
For any organisation building or deploying AI-driven land or climate monitoring, the practical standard the research points to is consent and shared authority, not just accuracy. That means involving Indigenous communities in decisions about what gets measured, how the results get categorised, and who has access to the resulting data, rather than treating community input as an afterthought once a model is already built. Google’s own AlphaEarth documentation describes over 50 partner organisations already using the dataset for conservation work, including groups like MapBiomas in Brazil, but the framework the CARE principles paper lays out suggests that partnership at the organisational level does not automatically guarantee the communities whose land is being mapped had any input into the process.
The Mexico and Colombia Precedents Worth Knowing
Two legal precedents show what a more consent-based approach can look like in practice. Mexico’s Federal Law for the Protection of the Cultural Heritage of Indigenous and Afro-Mexican Peoples and Communities, passed in January 2022, extends legal protection to community intellectual property and cultural expressions, framing cultural and land-related data as something communities have ownership over rather than a resource available for open collection. Colombia has taken a similar path since its 1991 constitution, recognising Indigenous authority over ethnic and cultural resources including archaeological and communal land data. Neither law was written with AI specifically in mind, but both establish the legal principle that land-related data about Indigenous communities is not automatically public or freely reusable, a principle AI-driven mapping tools have generally not been built to respect by default.
The Difference Between Mapping Land and Honouring It
The distinction researchers keep returning to is between land treated as a dataset to be optimised and land treated as an ongoing relationship a community has responsibility for. AlphaEarth’s technical capability, generating consistent, detailed maps of the same location over time, is genuinely useful for detecting deforestation, tracking agricultural change, or supporting conservation planning. Whether that capability serves Indigenous communities or bypasses them entirely depends on decisions made well outside the model itself: who gets consulted before a monitoring project launches, who controls access to the resulting data, and who benefits when that data eventually informs a policy or funding decision.
None of this argues against using AI for climate and conservation work. It argues for building consent and shared authority into that work from the start, rather than treating Indigenous input as a step to add once the technology already exists. As our own reporting on AI’s invisible infrastructure has explored, the tools that decide what gets measured and how it gets categorised carry real power, whether or not that power is visible to the people affected by it. For AI tools mapping the planet’s most sensitive and contested land, that power comes with a genuine obligation to the people who have been protecting that land long before any satellite existed to photograph it.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.