Former EPA officials tally 30 rollbacks speeding polluting AI data center buildout

Former EPA staff catalogue the rollbacks

According to The Verge, former officials of the US Environmental Protection Agency warned this week that the Trump administration is steadily dismantling environmental safeguards to speed up AI data center construction, and that the resulting pollution threatens people who may never live near a data center.

The warnings came from the Environmental Protection Network (EPN), an independent group of agency alumni, in a briefing and a new report. The document lists 30 federal actions since January 2025 that the authors say worsen health risks tied to data center pollution; 17 of them name AI or target data centers directly. The changes include efforts to curtail renewable energy projects, softened pollution limits, and exemptions for existing power plants and factories from certain standards.

The report follows early-term declarations by EPA administrator Lee Zeldin that deregulation would help make the country the world's "AI capital." Since then, The Verge reports, the agency and the White House have rolled back dozens of rules while cutting EPA staff and funding.

Policy groundwork for a fossil-fueled buildout

In July 2025 the White House published an AI Action Plan recommending streamlined or reduced regulation under the Clean Air Act, the Clean Water Act and Superfund law to accelerate permitting for data centers and chip factories. Tyson Slocum, energy program director at the consumer group Public Citizen, told The Verge at the time that the plan effectively uses emergency powers to grant sweeping exemptions for data centers and the fossil fuel infrastructure behind them.

The consequences, as EPN describes them, extend across the supply chain: policy support for new gas projects ranging from turbines installed on-site to power plants built specifically to serve tech companies, and renewed production of persistent "forever chemicals" driven by demand for advanced chips. Those chemicals break down slowly and accumulate in the environment and in human bodies.

Lynn Goldman, a pediatrician and former EPA assistant administrator, said at the briefing that a construction wave of this magnitude, billed as a revolution, should have prompted more vigilance from the agency, not less, and that the EPA is instead putting industry's needs first while weakening or reconsidering limits on toxic chemicals and pollutants. Her point about geography is central to the report: because the pollution comes from the power plants and factories supporting data centers rather than only the facilities themselves, exposure is not limited to neighboring communities.

The projected toll

The report cites research from UC Riverside, Caltech and the Rochester Institute of Technology estimating that air pollution linked to AI could cause up to 1,300 premature deaths and more than $20 billion in public health costs by 2028. EPN argues the total could climb higher once the 30 policy changes are factored in.

The group's proposed remedy is a "Data Center Health Protection Pledge": restoring clean air and water standards, rebuilding federal agencies' capacity to enforce those rules, and publicly reporting how the buildout changes pollution, who is being exposed and what it means for their health. Marc Boom, EPN's senior director of public affairs, said at the briefing that a more polluting AI buildout is not inevitable and depends on the choices government makes now.

The friction inside the agency is not new. As The Verge notes, hundreds of current EPA employees signed a public letter to Zeldin last year arguing that under his leadership the agency would increase rather than reduce risks to public health and safety, particularly around hazardous chemicals and drinking water.

In an emailed response to The Verge, EPA spokesperson Cora Mandy said each action the agency has taken returned regulations to the best reading of the Clean Air Act after years of overreach by previous administrations, and that the EPA remains committed to protecting human health and the environment while also prioritizing the goal of making America the AI capital of the world.

Why it matters

The physical footprint of generative AI — data centers, dedicated gas plants and chip fabrication — is being permitted under rules explicitly loosened to accommodate it. If the cited projections hold, the cost of faster AI deployment will be paid in measurable health outcomes and public spending, spread across regions far from any server hall. The dispute is not really about whether the buildout happens, but about whether anyone is monitoring its emissions and who bears the consequences when nobody does. For the infrastructure industry, it also signals that energy sourcing and siting decisions made now — gas versus renewables, on-site generation versus grid power — will carry regulatory and public-health scrutiny that could return in force if enforcement priorities shift.