新报告显示：84%的勒索软件攻击与人工智能有关

AI linked to 84% of ransomware attacks, new report finds

October 14, 2026 / 6:00 AM EDT / CBS News

Cybercriminals exploited artificial intelligence in ransomware attacks experienced by 84% of surveyed victims, according to a sweeping new global cybersecurity report, while 44% reported that attackers exploited their own AI systems.

The 2026 Ransomware Risk Report by cybersecurity firm Semperis polled 1,200 IT and security professionals across nine countries and eight industry sectors, revealing an evolved picture of cybercriminal enterprise—one that increasingly leverages AI to magnify its technical capabilities and escalate pressure on targets.

But the findings also underscored ransomware’s human toll. According to the report, more than a third of surveyed victims faced psychological or physical threats against employees, and 14% of surveyed organizations even reported employees or family members forced to change jobs, office locations or residences as a result of sustained ransomware threats.

In all, 69% of surveyed ransomware victims experienced multiple successful attacks sometime in the previous year, with 59% of respondents paying a ransom at least once, and 87% experienced attacks targeting their identity systems, or the digital infrastructure that controls access to their networks.

For Chris Inglis, former White House National Cyber Director under former President Joe Biden, artificial intelligence represents a turning point in the capabilities ripe for cybercriminals’ exploit —and he warns that a particularly consequential development is already underway: autonomous ransomware attacks.

“AI actually increases the capacity, scope, scale of attackers in understanding what perhaps might be the weaknesses in your system or perhaps increase the sophistication of their approach to you if they’re trying to get you to do something you shouldn’t do, that gives them an advantage, lets them in,” Inglis said in an interview about the report.

The cybersecurity expert and Semperis board member previously spent three decades watching technologies mature at the National Security Agency watching as attacks required substantially less direct human involvement – no longer simply a theoretical possibility.

“AI is an inflection point in their capacities to hold the same things at risk,” Inglis said. He later added, “I think we’re already seeing autonomous attacks. Autonomous ransomware attacks are included in the mix.”

Inglis described how a person lacking the technical expertise to penetrate an organization’s networks could turn to AI to write the necessary code, effectively instructing an AI system to target a company and hold its data hostage without a sophisticated understanding of how digital infrastructure operates.

“We have democratized AI to the point where even low level models can allow that to happen,” Inglis said, noting that the technology compounds the number of criminals capable of launching more technical attacks, while inviting experienced cybercriminals to penetrate networks more efficiently and on a much greater scale.

AI is changing the ransomware playbook

For years, criminals have relied on a familiar playbook: exploiting vulnerabilities, penetrating corporate networks, encrypting data and demanding a ransom payment in exchange for restored access or withholding sensitive information. Now, in more than eight out of 10 attacks, AI has been incorporated into that process.

AI has also enabled attackers to quickly identify weaknesses, tailor social engineering schemes to specific employees, automate an intrusion and readjust tactics as defenders try to react, according to Inglis.

“What’s not changed is the things that people or organizations had that is of value that they would pay money or time or energy to protect,” Inglis said. “And those things are still the object of a broad swath of criminals attention. They want to use whatever tools they can to try to hold those at risk so they can get you to pay them money, sometimes time, but really it’s about money.”

By lowering technical barriers, AI risks enabling a wider range of actors to attempt intrusions that previously demanded a wide swath of resources and more specialized training or know-how.

Tomer Bar, Semperis’s associate vice president of security research, warns that AI could enable ransomware developers to produce more sophisticated malicious code, reduce development time and make investigations more challenging.

“We’re starting to see more attacks with more rapid development, and attribution of these attacks will be more difficult for law enforcement,” Bar says in the report.

“We’ll also be less likely to see mistakes in attack encryption algorithms because attackers can test their code with large language models.”

When a company’s own AI becomes a weapon against it

One of the report’s most striking revelations is that 44% of surveyed victims experienced ransomware attacks exploiting their own AI agents or systems, a finding that raises an uncomfortable question for businesses and government agencies racing to integrate the technology into its networks.

Courtney Guss, Semperis’s associate vice president of crisis management, warns that autonomous AI agents can create security risks when they seek additional permissions or interact with sensitive corporate systems.

“If your company is adopting AI agents, you should consider that at some point, those agents will try to compromise your identity systems — even if that action isn’t meant to be malicious,” Guss writes in the report.

“That’s just how they operate. AI agents typically try to accomplish their assigned tasks by any means they can.” Guss continued. “One way they can gain more ability to operate in the environment is by getting permissions to other systems. The fastest path to get those permissions is to compromise your identity system, which effectively puts the agent into ‘god mode,’ enabling it to grant itself whichever permissions it needs to accomplish its task. Unfortunately, the agent then becomes a ticking time bomb.”

Unlike conventional software operating, some AI agents possess the ability to use multiple applications at once and make educated decisions on behalf of their users.

If an attacker manipulates just one of those agents—or compromises the credentials and permissions it relies upon—the attacker gains another way of moving deeper into an organization’s systems.

Still, Inglis cautions against assumptions that AI necessarily makes an organization more vulnerable to outside intruders.

“Many attackers already do what’s called living off the land – they’ll move into your digital infrastructure, onto your property or digital property, and they’ll use the tools that you have there or the insights or the data set,” Inglis said. “To blend in, to navigate the landscape, they try not to stand out, and they do that by using your tools. If AI is one of your tools, they’ll use those, but the weakness that let them in is probably the same weakness that existed before.”

Inglis argues that the fundamental vulnerability often lies within a company’s failure to control access to its networks or digital infrastructure in the first place. Rather than abandon AI out of fear, Inglis argues that organizations must embrace its defensive potential.

“I would turn that around. Flip the script,” he said. “You must use AI in order to counter that threat, and you must use AI for affirmative purposes.”

Julia Kimani-Burnham contributed to this report.

This story published first on CBS News at 6:00 a.m. on Oct. 14, 2026.