AI news, problem solutions & tool picks
DigestX

Problem Solutions

The AI-fueled ransomware boom is mostly a boom in claims

· 2 min read

Photo via Pexels

Claimed ransomware attacks spiked over the summer, WIRED’s Kernel Panic newsletter reported on November 14 — nearly 3,000 alleged attacks posted on leak sites, a 20–40% rise since spring. The numbers are real, and the word to watch is “claimed”: these are announcements on leak sites, not confirmed intrusions, and only 9.4% of the victims in one tracker’s count confirmed the attack themselves.

What the trackers actually counted

Three independent trackers tell the same story with different totals. Halcyon logged 2,766 claimed victims in the third quarter, up 39% on the second, spread across 111 groups in 122 countries, with more than 1,000 claims in August alone — its highest monthly total on record. ReliaQuest counted roughly 2,928 claims, up 30%. Comparitech counted 2,627, up 29% on the quarter and 61% year on year. But the second quarter was itself down 15% from the first, so the third quarter sits only about 10% above where the year started — and ReliaQuest writes in the same report that the rise “is unlikely to reflect” a real increase in intrusions. Forty-three first-time groups accounted for 93% of the growth. One new group, Global Secret Group, claimed 48 victims, more than half listed on a single day; another, Storm, listed 71 victims in 20 days. The Gentlemen led the quarter with 398 claims.

Where AI actually shows up

Fully AI-powered ransomware has not emerged, Recorded Future’s Allan Liska told WIRED — the successful groups are still largely human-operated. What AI is doing is lowering the cost of entry: language models write better phishing emails and handle ransom negotiations, and inexperienced operators are branching out on their own instead of working through ransomware-as-a-service providers. One new group, ReliaQuest found, is run by an inexperienced teenager who made more than $800,000 in under two weeks. Halcyon’s report documents the one place automation visibly scaled: AI agents harvesting more than 2,100 Azure AD token sets across 40-plus tenants in roughly 34 hours. Of 405 AI-linked malware samples Halcyon tracked this year, 12 reached a protected endpoint — and all 12 were detected. Halcyon’s Cynthia Kaiser, a former FBI cyber deputy director, said the US was comparatively steady while international attacks rose sharply, and that the rise is “in part fueled by AI”.

What to do about it

Treat leak-site counts as claims, not a threat metric — and assume the next wave starts with stolen logins, not malware. Patch and segment, keep offline backups, and watch for credential harvesting: the quarter’s clearest AI success was collecting tokens at scale. Recorded Future expects attacks involving actual encryption to fall as lone actors shift to pure data extortion — which means the ransom note may arrive without any encryption at all.