Indigenous leaders at the UN are grappling with a paradox: AI can help defend forests from illegal logging and mining, but the data centers running it depend on water, energy, and minerals pulled from Indigenous lands.
Indigenous communities are using artificial intelligence to catch illegal loggers, forecast droughts, and defend some of the planet's most intact ecosystems — even as the data centers powering those same tools strain water supplies and electricity grids in other Indigenous territories. That contradiction sits at the center of a new warning from United Nations experts, as reported by Grist.
The conventional pitch around AI and conservation treats the technology as a neutral upgrade: better satellites, faster pattern recognition, cleaner forests. What Indigenous leaders are pointing out is messier. The same infrastructure that helps a community in the Amazon spot a bulldozer from orbit is built on minerals, water, and energy often pulled from other Indigenous lands.
A new study by Hindou Oumarou Ibrahim lays out both sides. AI can monitor biodiversity, detect illegal mining, and predict climate impacts when combined with Indigenous knowledge. It can also accelerate land-grabbing, water overexploitation, and degradation driven by the sector's appetite for power and critical minerals.
The examples on the ground are specific. In Brazil's Acre state, agroforestry agents in the Katukina/Kaxinawá Indigenous Reserve are using drones, GPS devices, and AI forecasting to track invaders. A tool built by Microsoft and the Brazilian nonprofit Imazon ranks the reserve among the top five in the state for deforestation risk. In Nunavut, Inuit communities are pairing traditional knowledge with predictive models to find fishing grounds shifting under climate change. In Chad, Indigenous pastoralists combine participatory mapping with satellite data to anticipate droughts and protect transhumance corridors.
Each of those wins depends on infrastructure built somewhere else — and that is where the ledger flips. The satellites, servers, and training runs that make Indigenous-led monitoring possible draw on a supply chain that lands hardest on other Indigenous communities. Residents in Thailand's Chonburi and Rayong provinces have raised fears about water shortages and air pollution linked to data center expansion. Similar fights are playing out in eastern Pennsylvania and in Querétaro, Mexico. A single hyperscale data center can consume up to 5 million gallons of water per day for cooling and draw 100 megawatts or more of electricity — enough to power tens of thousands of homes — and the minerals feeding AI hardware are often mined near Indigenous territories.
Advocates point out that AI's environmental footprint includes significant water and energy consumption, as well as mineral extraction often occurring near Indigenous territories. The study also flags risks beyond extraction: drones and mapping tools deployed without consent can expose sacred sites, and Indigenous communities frequently lack the legal infrastructure to defend digital rights.
Kate Finn told Grist the consistent ask from Indigenous peoples is that free, prior, and informed consent be respected before data centers are built on their land. Lars Ailo Bongo, who leads the Sámi AI Lab at UiT The Arctic University in Norway, noted that progress is bottlenecked less by talent than by funding — Sámi developers exist, but the institutional budgets to build Sámi-aligned models do not.
The story here is not about whether AI belongs in conservation. It already does. The question is who sets the terms. When Indigenous communities run the tools, retain their data, and direct the funding, satellite imagery becomes an extension of stewardship that predates any algorithm. When the tools are imposed from outside, powered by extraction happening somewhere else, AI becomes another round of a very old pattern wearing new branding. The difference is governance, not gadgetry.
