The Risk AI Poses to Public Health

Artificial intelligence is making our systems more efficient, bettering our programming, making leaps for personalized medicine, and creating a new age of technology that will bring unprecedented advances to science. But underneath all its splashy innovation is a growing public health risk – a risk to the resources we unequivocally need to survive: water and air.

AI systems require the use of data centers, warehouses that house large computers that provide the computing power and data processing for LLMs, generative AI, among others.

AI systems need lots of water. The main function of water is to cool down the computers in data centers and to generate the electricity used to power them. However, the catch is that the water used for cooling has to be fresh. The model that first created ChatGPT, GPT-3, used 700,000 litres of fresh water in just its pretraining phase. Fresh water makes up just 0.5% of all the water on Earth, and many data centers are located in areas that rely on fresh water. Water used in data centers cannot easily be recycled. By 2027, AI may use 6.6 billion cubic meters of water globally; this is more than half of the UK’s total water usage.

In areas where there is plenty of water to be had, excess usage of water may be less risky. However, tech giants like Amazon, Microsoft, and Google are building and developing data centers in dry areas, it’s good for the data center, but incredibly detrimental to these areas and harmful to the local communities. AI companies generally keep the location of these data centers secret, but it’s no secret that as the AI boom continues, hundreds more data centers will be built across all continents. While companies claim to mitigate water usage, the secretive nature of private industries means that transparent data about water usage is not widely available.

It’s easy to predict the obvious environmental impact that this would have on the areas around data centers, for example, desertification, the collapse of ecosystems and significant decreases in biodiversity. However, it will also have an impact on the health of the residents in the surrounding communities.

Data centers emit pollutants such as fine particles, sulfur dioxide, and nitrogen dioxide generated from manufacturing servers, electricity generation, and the burning of fossil fuels. These pollutants can burrow deep into the lungs or bloodstreams and cause asthma, lung cancer, cognitive decline, or death. Fine particles emitted from data centers can travel large distances, and as there is no safe level of exposure, people who live near data centers are at a huge risk. Because these pollutants are airborne, pollutants generated in Florida could reach as far as Virginia, meaning that even more distant communities are at risk. In addition to air pollution, using fresh water in already scarce areas will mean that the local water is unsafe to drink. This means that areas surrounding the data centers face serious water scarcity problems and will experience a significant decline in clean drinking water as data centers continue to be used and built. AI is predicted to bring the public health burden to $20 billion in 2028, an impact that will be disproportionately felt in low-income or disadvantaged communities.

It is possible to mitigate the impact of AI on public health, but it requires transparency and

standardization regarding pollution caused by data centers. Sustainability reports should include clear disclosure on current and future water usage, and critical data on air pollutants emitted. Technology companies should contribute to easing the health burdens placed on surrounding communities and adjust their resources and usage accordingly. More attention should be placed on the unequal effect on poorer communities.

While AI may be the future of technology, it’s important to remember that public health should not be taken for granted. Ethical considerations must be taken into consideration as we propel forward a technology that has the potential to destroy the Earth’s environment and cause disease. Questions have to be raised. How can we balance AI’s undoubtedly positive applications with the impacts on community health? How do we regulate this groundbreaking new technology? And are we taking a gamble with people’s lives?