When AI researcher Jacob Coxon warned on X that OpenAI and Anthropic are “gambling with our lives,” he was talking about the threat of “self-improving superintelligence,” technology that can hack anything and “acquire real power.”
But that’s not the only threat from the AI boom.
Our rapid development of AI is already threatening humans when it comes to our health, due to the environmental impacts of AI data centers and the surge in fossil fuels to power them.
Already, AI is threatening billions of people’s access to natural resources, according to United Nations University.
By 2030, AI data centers’ water usage will match the needs of 1.3 billion people, that report found. Their power needs will be triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria, which encompass more than 650 billion people.
And a large portion of those data centers will be powered by fossil fuels, the main driver of planet-warming greenhouse gas emissions.
“In the absence of standards, the climate, health and environmental impacts of AI data centers are already being seen, with stark and alarming projections should data center growth continue unmitigated,” says Maria Fernanda Chavez, senior analyst for the Climate and Energy program at the Union of Concerned Scientists.
“Irresponsible management of AI development could spike heat-trapping emissions and push climate goals out of reach,” she adds, “backsliding on targets that have real consequences for the lives and livelihoods of people around the world.”
Big Tech is all in
As tech companies like Google and Microsoft go all in on AI, they’ve already reported an increase in their corporate emissions (and gotten quieter about climate change).
Microsoft’s emissions rose 25% in 2025 as it ramped up its AI data centers, and just one Microsoft natural gas-powered data center project could cause its emissions to skyrocket by 44%
These emissions exacerbate climate change, which threatens the entire planet, especially through deadly disasters like heat waves, floods, and storms.
But infrastructure that causes those emissions has other localized impacts, too, like air pollution.
“The health costs of how we choose to power AI are not hypothetical, and they are not decades away,” says Michael Cork, an environmental health scientist and founder and CEO of EmPower Analytics Group, which measures the health impacts of data center growth.
In order to meet the growing demand for electricity, data centers are driving up demand for new gas power, on-site gas turbines, and diesel generation.
In the U.S., the AI boom has already driven so much development of new gas power that we’re outpacing China, with nearly three times the amount of gas-fired capacity in the works.
“The air pollution from those sources is the same pollution linked to premature deaths, heart attacks, and asthma attacks in children, with nearby communities often bearing a disproportionate share of the impact,” Cork adds.
“Measurable health consequences”
Coxon’s concerns about AI prompted a public discussion about if we should be regulating such companies. It’s not just AI safety regulations that are needed, experts say, but also regulations around these health impacts.
The topic of air pollution, Cork says, is “largely being left out of decisions about where and how [data center] infrastructure gets built.”
But it will have—and is already having—real consequences.
“If we’re going to have a serious conversation about whether AI could someday threaten humanity, we should also be willing to confront the measurable health consequences of how we’re powering it today,” Cork says. “Understanding those consequences isn’t an obstacle to progress. It’s the minimum we should expect.”
