Artificial intelligence is changing environmental management — sometimes in the least glamorous places.
In the UK’s South West, a new AI system developed by South West Water (SWW) analyses data from 12,000 sensors embedded in the sewer network, spotting changes in water levels that signal potential blockages. According to the BBC, the company estimates that this early-warning system has already prevented around 200 pollution incidents.
By learning how water levels rise and fall, the algorithm can predict when a sewer might overflow, giving engineers time to intervene. The system, says SWW’s Helen Dobby, effectively gives the utility “12,000 extra pairs of eyes on our network.”
The same technology is also being used to scan CCTV footage of underground infrastructure, automating the identification of leaks, cracks, and fat build-ups — the tedious but essential maintenance work that prevents environmental damage.
It’s a striking example of AI’s potential as a guardian of the unseen environment: invisible, efficient, and untiring.
But while AI is helping utilities like SWW monitor water quality, the technology that enables it is itself consuming vast quantities of water and energy.
A separate BBC Scotland investigation found that data centres powering AI systems already use enough water in Scotland alone to fill 27 million half-litre bottles a year — a figure that has quadrupled since 2021.
These facilities — the warehouses of high-performance computing that run everything from ChatGPT to Google’s Gemini — need enormous cooling systems to prevent overheating. The majority currently use “open-loop” designs that rely on a continuous supply of mains tap water, though developers are beginning to adopt more sustainable closed-loop or wastewater-fed alternatives.
Scottish Water’s Colin Lindsay called the increase in consumption “significant,” warning that as the country welcomes new AI industrial parks, “supplying them all with tap water would be a real concern.”
Together, these stories illustrate the double-edged nature of AI’s environmental role.In one context, it’s a tool for precision management — detecting and preventing pollution.In another, it’s a resource-intensive industrial force driving new forms of ecological strain.
As Professor Ana Basiri of the University of Glasgow told the BBC, the water and carbon footprint of AI data centres is “very significant,” equivalent to each person in Scotland driving up to 90 extra miles every year.
Basiri argued that the issue is largely invisible — hidden in server racks and cooling systems rather than visible pollution — but no less real. “There is a huge amount of carbon dioxide emissions and water use related to data centres that we often forget about,” she said.
The same algorithms that promise efficiency and environmental foresight also require energy-intensive infrastructure that demands constant cooling, vast amounts of electricity, and now, increasingly scarce water.
The tension between AI’s benefits and burdens is becoming one of the defining questions of sustainable innovation.
AI excels at monitoring and predicting — whether that’s floods, leaks, or wildlife migration — but the computational power behind it carries environmental costs that are harder to monitor, especially when data centres are privately owned and opaque about their consumption.
SWW’s trial, developed with researchers at the University of Exeter, offers a glimpse of how AI can lighten humanity’s footprint by preventing waste and damage. But Scotland’s data centre expansion reveals how the same technology, at industrial scale, can deepen it through energy and water demand.
The Scottish government has acknowledged this trade-off, encouraging developers to adopt “closed-loop” water systems and use treated wastewater rather than drinking supplies. Meanwhile, OpenAI — the company behind ChatGPT — told the BBC it is pursuing “water-positive” projects and building renewable-powered data centres in Norway.
For now, AI’s environmental story remains split between the micro and the massive.
In Devon, a small team of engineers uses machine learning to stop fatbergs.
In Scotland, multinational firms build data parks the size of football stadiums to train billion-parameter models.
Both are parts of the same ecosystem — one local and life-saving, the other global and resource-intensive.
The challenge for regulators, governments, and AI developers alike is to close the loop — to ensure that the intelligence managing environmental systems doesn’t become another system in need of environmental management.
If South West Water’s AI project shows how digital oversight can reduce waste, Scotland’s data centre dilemma reminds us that no technology is inherently sustainable.
AI’s environmental value will depend not on its intelligence, but on the infrastructure, energy sources, and accountability frameworks built around it.
AI is often described as the most transformative technology of the century — and it may well be.
But as the stories from the BBC show, transformation cuts both ways.
Whether AI becomes a cornerstone of environmental stewardship or another driver of ecological strain will depend on choices made now — about transparency, design, and resource use.
In the end, the intelligence we need most may not be artificial at all, but collective: learning to align digital progress with planetary limits.