AI Is Expanding the Ransomware Field, While Human Operators Still Lead
- Cargo
- Machine Learning
- Shipped
- Shipper
- Matt Burgess, Lily Hay Newman
- Readers
- 608
AI is helping more attackers launch ransomware operations and improve phishing, while established groups still rely on people to run attacks.

AI is making ransomware operations easier to start and some parts of an attack easier to carry out. Researchers have seen attackers use large language models to improve phishing and handle negotiations, while automation can help with tasks such as checking code and finding ways into systems. The operations making the most impact, however, are still largely run by people.
That distinction matters. Researchers have not seen fully AI-powered ransomware emerge, and the existing criminal business model already works. But tools that lower the skill and effort needed for specific tasks can help more operators try their hand at extortion. That may add to the volume of attacks even if the underlying methods remain familiar.
Reported attacks are rising, but the count is uncertain
There is no precise count of ransomware incidents. Some victims never disclose attacks, while groups can exaggerate or fabricate claims on leak sites, where they name supposed victims. Researchers who track those claims and try to remove duplicates say the reported volume rose over the summer. Some estimates put the number of alleged attacks at nearly 3,000, up 20 to 40 percent from spring.
Those figures describe claims, not verified compromises. Still, researchers say the rise sits on top of a high level of ransomware activity that has persisted for years. Cynthia Kaiser, who leads research at anti-ransomware platform Halcyon, says the latest increase is most visible outside the United States. US activity remained high but steadier, while international attacks grew. Kaiser believes AI is one factor behind that change.
The connection may begin before ransomware reaches a victim’s systems. Attackers often use phishing and social engineering to gain access, and AI can help them produce more convincing messages. Kaiser points to a recent rise in successful scams in Japan, where attackers have used AI to generate persuasive Japanese-language content. The same ability could support ransomware operators seeking an initial foothold.
AI may be lowering the barrier to starting a group
Several operations, including the Gentlemen, Qilin, and Akira, have remained prominent in recent months. Alongside them, the field has become more fragmented, with new groups making rapid claims. Halcyon’s analysis counted about 111 groups claiming attacks between July and September.
Two examples illustrate how quickly those claims can accumulate. ReliaQuest reported that Global Secret Group claimed 48 victims, listing more than half on the same day. Another group, Storm, listed 71 victims in 20 days. Those figures have not all been verified. Even false or inflated claims can leave businesses unsure whether they have been targeted or what access an attacker may have.
Brandon Tirado, ReliaQuest’s director of threat research, sees AI as one reason more operators may be able to start their own groups. Instead of relying on established ransomware-as-a-service providers, some are operating independently. Tirado says his team identified an inexperienced teenager behind one new group who made more than $800,000 in under two weeks. He did not name the individual.
Kaiser says AI can help newer operators deceive targets, exploit vulnerabilities, review their code, and try to gain a second route into a victim’s systems by compromising remote-access tools. Those capabilities can help an operator attempt more steps without the experience or support that a mature group might have. But it is still unclear how many new operations can sustain that pace. Kaiser expects some to surge and then fade as they encounter the demands of running an operation over time.
More extortion may not mean more encryption
Established groups often combine two forms of pressure: they encrypt a victim’s systems and copy data, then threaten to publish that information. This approach is commonly called double extortion. Independent operators may find data theft and publication threats simpler to carry out than deploying ransomware that encrypts systems.
That shift complicates how attack counts should be read. Allan Liska, a ransomware researcher and threat intelligence analyst at Recorded Future, says the number of attacks his team has seen this year already exceeds the total for 2025. Yet attacks involving encryption may decline year over year. An increase in extortion claims therefore does not necessarily mean more organizations are dealing with encrypted systems; stolen data and threats to disclose it can drive the count as well.
Liska says human operators remain central to successful ransomware groups, even as they experiment with AI for phishing and negotiations. His caution is useful: better tools can speed up particular tasks, but they do not by themselves show whether a group can select targets, maintain access, manage extortion, and keep operating. For defenders, the near-term signal is a wider pool of people attempting attacks and more polished social engineering. Tracking both data extortion and encryption will help distinguish a rise in claims from a change in how victims are actually being pressured.