Ransomware groups increasingly use LLMs for phishing and negotiation, number of attacks rose in summer 2026
According to researchers from Recorded Future and Halcyon, the number of ransomware attacks reported on leak sites rose to roughly 3000 in the summer of 2026, up 20–40% since spring. They say attackers are increasingly using LLMs to create phishing emails and negotiate with victims, but fully AI-driven ransomware has not yet emerged.
Researchers tracking ransomware attacks (Recorded Future, Halcyon, ReliaQuest) report that the number of attacks published on leak sites reached roughly 3000 in the summer of 2026, which according to some estimates represents an increase of 20 to 40% since spring. According to Cynthia Kaiser of Halcyon, the increase is driven primarily by international attacks, while the number of attacks in the US has remained relatively stable, albeit high. Between July and September 2026, according to Halcyon's analysis, roughly 111 ransomware groups claimed attacks.
According to researchers, LLMs are so far mainly used in attacks for partial tasks — creating more convincing phishing emails and negotiating with victims — while the attacks themselves are still directed by humans. Allan Liska of Recorded Future points out that groups are experimenting intensively with AI, which he says could change quickly. As an example of a broader trend, researchers mention a rise in fraud in Japan, where AI for the first time allows foreign attackers to generate convincing text in Japanese.
Alongside established groups like Gentlemen, Qilin and Akira, a number of new and smaller actors are reportedly emerging, according to the source. The new group Global Secret Group reported 48 victims according to ReliaQuest's analysis, while the group Storm reported 71 victims in 20 days. Brandon Tirado of ReliaQuest claims that one of the new groups is run by an inexperienced teenager who earned over $800 000 in less than two weeks. According to Kaiser, AI allows more individuals to set up their own groups without relying on traditional ransomware-as-a-service providers, although it is not clear how long these new groups will last. Liska adds that, according to Recorded Future's data, this year's number of attacks already exceeds the whole of 2025, though the number of attacks involving actual data encryption may be declining year-on-year in favor of attacks focused only on data exfiltration and extortion.
Why it matters
The rise in attacks and the easier entry of new, less experienced actors into the ransomware business increases the likelihood that even a smaller company, which previously was not a typical target, will encounter an attack attempt or convincing phishing. More convincing AI-generated phishing emails, now also available in languages other than English, reduce the reliability of previously used warning signs of fraud (grammatical errors, clumsy text).
Two audiences, two different impacts
What this means
For individuals
Phishing emails generated using AI are of higher linguistic quality and harder to recognize based on previous warning signs, such as grammatical errors.
For a business
The rising number of ransomware groups and the increasing quality of phishing emails and negotiation messages generated by LLMs increases the risk of a successful initial breach of a company network and requires a reassessment of employee training and phishing detection processes.
Risks and complianceCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.