AI attacks are far faster than human defenses—how can America's power grid, banks, and hospitals hold the security bottom line?

Anne Neuberger, a16z's global affairs lead and former U.S. deputy national security advisor, proposed a public-private national cyber defense plan in the Wall Street Journal, warning that AI is compressing the window between vulnerability discovery and exploitation far faster than human-speed defenders can patch legacy operational technology systems that have run physical equipment for over 20 years. The plan targets the few hundred core operators and shared suppliers whose failure would trigger nationwide cascading outages across five sectors—electric power, banking and payments including Fedwire and ACH, water, telecommunications and 911, and hospitals and health-record vendors—rather than replacing all aging equipment. It would place a CEO council under the White House's Super Intelligence Force, with industry CEOs signing commitments to zero internet exposure for control systems, 72-hour manual-operation drills, and time-bound vulnerability containment, reported through monthly ISAC scorecards. Existing Information Sharing and Analysis Centers would be converted into AI-driven execution engines, while cloud providers, telecom carriers, AI labs and control-system makers would supply scalable defenses and digital twins for testing fixes before deployment to live grids or water plants. The plan also calls for liability protections and "safe harbor" incentives for participating firms and for equipment makers—many foreign—to be required to act on disclosed vulnerabilities as a condition of selling to critical operators.