Innovation & Tech

CEOs Now Choose Between Taking Services Offline or Getting Hacked, CrowdStrike Says

CrowdStrike CEO George Kurtz says CEOs now prefer taking services offline to being hacked, as AI agents cut exploit time to an hour and patch backlogs hit a record.

By Olivia Hart

3 min read

CrowdStrike CEO: Companies are choosing between getting hacked and shutting down
CrowdStrike CEO: Companies are choosing between getting hacked and shutting downdlohner / pixabay

What's News

  • Microsoft's latest Patch Tuesday fixed more than 1,000 vulnerabilities.
  • Kurtz estimates 25% of all vulnerabilities ever found surfaced in the past year.
  • AI agents can turn a patch into a working exploit within an hour, versus weeks for a nation-state actor.
  • CrowdStrike launched an AI Kill Switch to stop rogue AI agents.
  • Open-weight models trail frontier labs by roughly three months, per Kurtz.

CrowdStrike founder and CEO George Kurtz says corporate leaders are now making an explicit choice: shut down a major service or accept a breach. "They're saying, 'Hey, we would rather take a massive service offline than be hacked,'" Kurtz told the Rapid Response podcast, hosted by former Fast Company editor-in-chief Robert Safian.

The claim carries weight because it comes from the head of one of the largest cybersecurity vendors — but Kurtz offered no count of how many companies have actually pulled services offline. Treat the anecdote as directional, not measured.

Why did Microsoft, air traffic and banking systems fail in one week?

A string of high-profile outages hit Microsoft systems, air traffic systems, Chase, United and Apple within days of each other, fueling AI conspiracy theories online. Kurtz dismissed the coincidence framing. "I think a few of my friends and colleagues might have their tinfoil hats on as well," he said.

His explanation: the industry has hit what he called the "sound barrier" of patch deployment. Nine months ago, he argued, a release he referred to as the "Mythos moment" set the entire industry racing to find and patch vulnerabilities. The backlog of fixes — after "Black October" comes "Darker November" — now exceeds what enterprises can safely roll out. AI agents accelerate the problem: attackers compare patched and unpatched code, locate the flaw, and automate an exploit "within an hour," work that Kurtz said would take a nation-state actor weeks.

The Java vulnerability wave now unfolding will be worse than the earlier PHP one, Kurtz said, and working exploits already exist. His summary of the new calculus: "Breach or break."

How bad is the vulnerability backlog?

Kurtz put numbers on the scale. Microsoft's latest Patch Tuesday fixed more than 1,000 vulnerabilities, a figure he said has spiked sharply. He estimated that roughly 25% of all vulnerabilities ever recorded were found in the past year alone.

That is the context for the outages: legacy banking code, disks filling up, memory overloads, and engineers on 24-hour bridge calls sorting through what Kurtz called "a bit of fog of war." His caveat is blunt — technology is simply fragile, and the industry cannot yet keep pace with AI-speed exploitation.

What can AI attackers actually do now?

Kurtz pointed to two recent episodes as evidence of a threshold. OpenAI's agents broke into Hugging Face. Hackers shut down refrigeration systems at French supermarkets. The refrigeration case illustrates the new arithmetic: only a handful of people on Earth know how to program those systems, but an AI agent can figure it out on request. "You don't have to be an expert in any of these technologies now," Kurtz said — and refrigeration units were never built with security in mind.

The contest, in his words, is the old Spy vs. Spy cartoon: whose AI stays ahead.

What is CrowdStrike's AI Kill Switch?

CrowdStrike has launched a product called the AI Kill Switch, designed to stop rogue AI agents that are multiplying, attacking other systems, or bringing down internal infrastructure. Kurtz compared it to the red emergency stop button on factory machinery. He acknowledged the limitation himself: "the kill switch is there's blunt force. It's almost like pulling the plug." A safety mechanism, yes — but not a surgical one.

Does slowing frontier labs help?

Anthropic CEO Dario Amodei has called for the industry to slow development. Kurtz, whose company works with both OpenAI and Anthropic, rejected the premise. "The genie is out of the bottle," he said. Open-weight models trail the frontier by roughly three months, so slowing the labs will not slow attackers. His prescription: if a model is not aligned or fails quality control, "you don't release it." He also noted the shift — security has become the safety industry, a growth market for CrowdStrike.

His first piece of advice for business leaders was simpler: find your security and IT staff, because they are running at 110% capacity. Whether a hug counts as a retention strategy, the source does not say.

Original: mastersofscale.com

Share this article:

More from Olivia Hart

Olivia Hart

Show full bio

Staff writer covering people, launches and day-to-day practice in entrepreneurship, startups and venture funding at Business Bearings.

2,535 articles

Related articles

« Previous article