Environmental organisations are putting AI to work on conservation problems while challenging the resource demands of the infrastructure behind it. Their choices show why adopting a useful application does not settle the wider question of sustainable deployment.
In a November 7 Reuters report, all 15 environmental, conservation and development organisations interviewed said they used AI in some form and were concerned about data centres’ environmental impacts. That describes the interviewees, rather than a representative survey of the entire sector.
The report identifies concrete attempts to limit unnecessary computation. The World Resources Institute told Reuters it restricts resource-intensive image and video generation and reuses cached query results. Greenpeace International said some of its AI use comes through existing software, while its internal guidance stresses transparency, human oversight and deliberate use.
Moving the work closer to the data
The Nature Conservancy’s own account describes a fisheries system that analyses monitoring video aboard longline vessels. Computer vision detects, tracks and classifies catch; daily summaries can then be compared with captains’ electronic logbooks to flag potential discrepancies before a vessel returns to port.
The organisation says experts remain responsible for validating predictions. Its description supports a specific monitoring workflow, rather than a claim that AI has replaced fisheries inspectors. Processing footage onboard also changes what needs to be transmitted, but the announcement alone does not establish the system’s net carbon or water savings.
In Cambodia, Conservation International documented satellite-triggered fire alerts reaching community patrols around Tonle Sap. Its June account describes people checking alerts and responding on the ground. This is background on an established programme, not a newly launched November service.
Count the infrastructure as well as the benefits
The broader demand is still growing. The IEA’s updated 2026 outlook projects global data-centre electricity consumption rising from 485 TWh in 2025 to about 950 TWh in 2030. Those figures cover data centres overall, not AI alone, and are projections rather than measured future consumption.
The practical question for an organisation is therefore specific: what task improves, how much computation it needs, and who checks the result. A successful conservation application can demonstrate operational value without, by itself, proving that its environmental benefits outweigh its full infrastructure footprint.
