Article

Axios

Axios

Behind the Curtain — AI 2027: The reckoning

9 hours ago

We've talked to more than 50 leaders in government, AI, business and cyber — from let-'er-rip AI accelerationists to shut-'er-down critics — and they were unanimous on one chilling assessment: Cyberattacks on business and critical infrastructure will be constant, increasingly sophisticated and highly disruptive over the next 18 months. These leaders describe 2027 as the year cyber becomes like weather: Outages will be frequent, disruptive, occasionally scary, and something people and companies must learn to navigate and live with daily. The chances of widespread disruption or real damage are rising fast. Why it matters: We write this neither to forecast nor frighten. Rather, we're sharing what government and AI leaders say privately but won't say publicly. The big picture: The disruptions of the past two weeks — from a global Microsoft outage to the FAA grounding air travel to a consumer banking glitch at Chase — were a mild preview of what experts expect is coming. Bridgewater Associates, the world's largest hedge fund , put it bluntly in an assessment this month: "[We] are managing money based on a one-in-three chance of an economically destabilizing attack within the next year." Horacio Rozanski , chairman and CEO of Booz Allen, calls this the "fog before the war" — a foretaste of greater chaos to come, when agentic invaders reach their full potential and can attack "at the speed of thought." "Now is the time to take as many precautions as you can," Rozanski said. "Recognize that this is coming." Most of these malfunctions appear to have resulted from companies and government trying to secure their computer and infrastructure systems against future large-scale cyberattacks. This pattern will persist. Many future ones will be actual cyberattacks, carried out by U.S. rivals or smaller bad actors, using increasingly powerful AI to spot and exploit security vulnerabilities across society and our daily lives. Most top AI and government leaders expect at least small-scale strikes and fear much larger ones. The best-case scenario described to us: regular, random outages and small-scale scares like losing bank access or air travel for a day. The reason: Every government agency and company will mobilize to upgrade their software systems. So errors are inevitable. These are assumptions based on private intelligence and an understanding of AI's capabilities, both to cause and fix glitches. What worries top officials most : With thin AI expertise, an understaffed government and major companies are trying to keep pace with a technology that improves daily. Even if government were loaded with competent experts, which it decidedly isn't, the sheer scale of vulnerabilities across federal, state and local government, not to mention all of business, would make it impossible to reliably secure. Think of it as whack-a-mole with infinite moles. Until this past month, the White House and Congress have focused largely on safety regulations and data centers, not imminent cyber risks. Now, even the White House's Super Intelligence Force ( SIF ) is scrambling to get its head around the rising risks while appraising near-daily reports of possible breaches. And President Trump just ousted the White House national cyber director and named Kirsten Davies, who was the Pentagon's chief information officer, as the replacement. Davies has deeper subject-matter expertise than her predecessor, former RNC official Sean Cairncross . Step back: Only by understanding the state of AI + cyber can government and business mobilize to prevent possible shutdowns across critical banking, electricity and water systems. This is our crack at distilling our conversations in a digestible way for Axios readers: The double-edged sword. The good news: America's frontier AI labs (Anthropic, OpenAI, Google, etc.) built the most powerful AI on the planet, especially for building software. The bad news: AI that's great at building software is equally great at finding — and exploiti