Watch: Ex-NSA Hacker Says AI 'Agent Swarm' Gutted A Company In 14 Minutes! - And Nobody Noticed For Hours
November 22, 2026
Source: Zero Hedge
In an interview that's both alarming and possibly more narrative shaping surrounding the dastardly dangers of AI (why not both?), former NSA hacker David Kennedy says an AI "agent swarm" broke into one of his clients, slipped past its two-factor authentication, stole data and encrypted its entire file system in 14 minutes - and he was still cleaning up the mess when he went on air Friday.

"This all took 14 minutes. 14 minutes. Never seen anything that fast before," Kennedy, founder of cybersecurity firm TrustedSec, told CNN's Erin Burnett, adding that the company "didn't even detect it for about four and a half hours."
Kennedy said the victim called TrustedSec Friday morning - "not one of the customers that we monitor 24-7" - and agreed to let him describe the basics without naming it "because they've never seen anything kind of ripped through their systems this fast." He said he almost canceled his appearance because he expected to spend the whole day on the incident.
Kennedy said the attack began with valid credentials. "We saw over 30 different types of agents probing their infrastructure at the exact same time," he said. The agents "logged in with their usernames and passwords, and were able to find a flaw in their application on how they use what's called two-factor authentication."
"It bypassed that. They got access to the system. They exfiltrated a substantial amount of data, and then they had a ransomware event," he said.
According to Kennedy, ransomware crews split the work: some groups "phish and harvest credentials, usernames and passwords from companies," then sell them to "more sophisticated groups that can actually launch attacks against the system."
"The overall message here is that, hey, this stuff is happening, and it's happening quicker than we can typically respond to it," he said.
Fourteen minutes is breakneck speed vs. what security firms were publishing only weeks ago. In September, Palo Alto Networks' Unit 42 described an intruder using AI agents to work through an enterprise network in roughly 10 hours or less, a job it estimated would take coordinated human red teams about two weeks. On Nov. 13, the FBI warned that ARTEX, a Chinese-built AI hacking framework, lets attackers go from first targeting to privileged access in under three hours, as we reported Thursday.
Kennedy's account comes as the FBI investigates a suspected breach of its own National Crime Information Center, the latest in a run of outages and disruptions at critical systems from the FAA to large US banks.
In the case Kennedy is working, the company only found out when its encrypted system fell over. "The system was still running, and eventually it errored out. The system restarted, and then they noticed they had an outage," he said. By the time it worked out what had hit it, about four and a half hours in, the attackers "were already gone for, you know, over four hours."


