ai cybersecurity nicole perlroth erin burnett outfront david kennedy
David Kennedy explains the speed and scale of a cybersecurity attack that only lasted 14 minutes Pete Linforth/Pixabay

A cybersecurity expert has warned that companies are running out of time to defend themselves against AI-powered attacks after a multi-agent swarm reportedly bypassed two-factor authentication, stole data and encrypted a customer's file system in just 14 minutes.

David Kennedy, a former NSA hacker, described the incident on CNN's Erin Burnett OutFront, saying the unnamed company did not identify the breach until its systems went offline hours later.

The warning cuts through debate over whether fears surrounding artificial intelligence are overstated. Kennedy said he was responding to an active incident involving one of his customers when he spoke to Burnett.

Nicole Perlroth: 'AI Is Turning up Vulnerabilities'

The account shared the interview as Kennedy explained the speed and scale of the attack. The post brought the incident into a wider discussion about AI-driven cyber threats.

Cybersecurity journalist Nicole Perlroth also pushed back against comparisons with the Y2K computer scare. In a post on X (formerly Twitter), she rejected the idea that the current moment was simply another technology panic, pointing instead to the reported breach and the speed at which the attackers moved.

Her warning was blunt. Organisations, she argued, have been 'living on borrowed' time after years of prioritising speed and growth, sometimes at the expense of security. AI is now making it easier for attackers to exploit weaknesses that may have remained hidden inside company systems for more than a decade.

'We have been living on borrowed time. AI is turning up vulnerabilities from a decade+ of "move fast and break things." We are on attackers' clocks now,' she wrote.

Neither social media post independently establishes the technical details of the breach. Those came from Kennedy, who said the customer had authorised him to discuss the basics of the incident but not reveal its identity.

AI Agent Swarm Took Just 14 Minutes, Kennedy Says

Kennedy told Burnett that more than 30 types of agents simultaneously probed the customer's infrastructure. The attackers used valid usernames and passwords before identifying a flaw in the way the company's application handled two-factor authentication, the additional security step intended to verify a user's identity.

After bypassing that protection, they gained access, exfiltrated a substantial amount of data and launched ransomware that encrypted the company's entire file system, Kennedy said. The sequence took just 14 minutes.

Kennedy said the company did not detect the incident for approximately four and a half hours. Its systems continued running after the encryption, before eventually producing errors and restarting. Staff then noticed an outage, by which point the attackers had already completed their operation.

Kennedy said he had considered cancelling his television appearance because he expected to spend the day working on the incident. He stepped away to explain what had happened, stressing that the broader concern was how quickly attackers could operate compared with the time defenders needed to respond.

He also outlined how some ransomware operations divide the work. Less sophisticated groups may harvest usernames and passwords through phishing before selling those credentials to more capable actors who can launch attacks against company systems.

The account raises an uncomfortable question for businesses that treat multi-factor authentication as a dependable last line of defence. In this case, the attackers reportedly found a way around it. Kennedy's description has not been independently verified, and the customer remains unnamed, meaning the precise vulnerability and full extent of the data theft are not publicly known.

Defenders may have hours to understand an intrusion that AI-driven attackers can reportedly carry out in minutes.