The AI doom debate is missing the grid’s real cyber problems

On July 22, a sudden grid disturbance in Ashburn, Virginia caused more than three gigawatts of data center load to disconnect at once — the largest such event in the history of PJM, the grid operator for a large swath of the mid-Atlantic. A faulty high-voltage line triggered the drop, but it highlighted a bigger worry among cybersecurity experts: as AI-driven data centers multiply, they create new, large-scale points of weakness for the electric grid.

Experts say artificial intelligence has made grid attacks easier to pull off. Tasks that once required teams with specialized skills, such as writing malware or understanding power-plant equipment, can now be done with AI assistance, lowering the bar for would-be attackers, including foreign actors. Iran-linked hackers have already targeted U.S. water utilities in multiple states. Meanwhile, much of the grid's control equipment predates modern cybersecurity practices, and older "walls" meant to isolate critical systems are easier to breach than before. Smaller utilities and newer renewable energy sites, which often rely on remote access and may have fewer cybersecurity requirements, are seen as especially exposed.

None of this points to a specific action for homeowners, but it underscores that grid reliability and security are shifting issues as data centers and renewable energy expand. Programs like Dragos's free Community Defense Program and a White House pilot called Project Watershed 250 aim to help smaller utilities improve their defenses, which could affect the reliability of power in some communities over time.