Proposed U.S. bill seeks to halt data center growth pending AI rules

U.S. Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez, who serves as the co-sponsor of the legislation, introduced companion bills in their respective chambers of Congress today. The proposed legislation seeks to halt the construction of new data centers that exceed a specific energy threshold, specifically targeting any new projects with peak power loads exceeding 20 megawatts. To clarify these terms, a data center is the physical infrastructure housing computing systems, while peak power loads refer to the maximum electrical demand of a facility. This threshold means the legislation would target the largest computing facilities that power modern artificial intelligence models. Under the terms of the bill, these infrastructure projects would be paused until the U.S. Congress enacts comprehensive regulation for artificial intelligence.

The legislative push draws on warnings from several tech luminaries regarding AI safety. The sponsors of the bill cited remarks from industry figures who have discussed their fears of AI and called for stricter rules or pauses on its development. These cited individuals include Google DeepMind chief Demis Hassabis, Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Nobel Prize-winner Geoffrey Hinton, who are all cited in the proposal as tech luminaries regarding AI safety. The proposal also highlights a warning from Elon Musk, who questioned the lack of regulatory oversight by stating, “AI is far more dangerous than nukes. So why do we have no regulatory oversight?”

This regulatory effort comes amid public apprehension in the U.S. regarding the development of artificial intelligence. A Pew Research poll conducted in March found that only 10% of Americans surveyed said their excitement about AI outweighed their concern. Despite this public sentiment, the bill might be seen as an opening bid for what U.S. AI regulation should look like. The proposed framework seeks to have the U.S. government review and certify models before release, establish protections against AI-driven job displacement, limit the environmental impact of data infrastructure, and require union labor for construction. It also aims to prohibit the export of advanced chips to countries without similar rules. However, the proposal faces significant hurdles to enactment. The combination of massive political spending by AI companies and geopolitical fears of losing an AI arms race with China may make the legislation difficult to enact.

Why it matters

This legislation represents a significant escalation in the backlash against AI infrastructure. It signals that the physical footprint of AI development is becoming a primary battleground for federal regulation in the U.S.