As artificial intelligence systems become more deeply embedded in society, researchers are raising alarms about the energy demands required to train and operate them—and the broader risks that may follow.
At the Institute for Trustworthy AI in Law & Society (TRAILS), experts are examining how AI systems interact not only with each other, but also with critical infrastructure and governance systems. These interactions, they warn, could introduce new vulnerabilities.
“One of the key concerns is just the massive amount of computing power it takes to train and operate these models,” said David Broniatowski, a professor of engineering management and systems engineering at George Washington University and deputy director of TRAILS.
Because of these immense resource requirements, only a handful of companies currently have the capacity to develop and deploy advanced AI systems at scale. That concentration of power, Broniatowski noted, gives those companies significant influence over how AI technologies are designed and implemented.
Beyond questions of control, the physical impact of AI infrastructure is also drawing scrutiny. High-performance computing systems consume vast amounts of electricity, generating significant heat. That heat, in turn, can affect local environments, including water quality in regions where cooling systems rely on large volumes of water.
Researchers are also considering how dependent modern society may become on these energy-intensive systems. In the event of a major natural disaster—such as a storm or grid failure—that disrupts power infrastructure, AI-dependent services could be compromised.
“What happens if we lose access to that computing power?” Broniatowski asked. “What are the consequences for our ability to keep society running if we rely on deep computation for critical functions?”
Ultimately, the concerns extend beyond energy use alone. TRAILS researchers are studying the full lifecycle of AI development, from engineering and deployment to governance and user interaction. Their goal is to better understand how AI systems integrate into—and potentially strain—the broader systems that support modern life.
As AI continues to evolve, these findings highlight the need for careful planning around energy use, infrastructure resilience, and equitable control over the technology shaping the future.
—Story by UMIACS communications group