Ransomware activity in Q1 2026 held at its second-highest level on record while internal AI adoption is emerging as a cyberrisk for companies, according to Travelers.
The first quarter of 2026 produced 2,405 ransomware victims posted to leak sites, a 2% decline from Q4 2025’s all-time high but a 7% increase over the same period a year earlier, according to Travelers’ Q1 2026 Cyber Threat Report.
The figures suggest that elevated ransomware activity, rather than receding after a record peak, has become the new baseline. Travelers’ internal claims data reinforces the longer-term trend: ransomware claims have increased 80% since 2022.
A More Fragmented Threat Landscape
The headline volume figure obscures a significant structural shift in who is carrying out the attacks. Eighty-four distinct ransomware groups were active in Q1 2026 — the highest count in Travelers’ dataset going back to 2020, up from 70 in Q1 2025 and 63 in Q4 2024. Nineteen of those groups made their first appearance on leak sites during the quarter, while 20 established groups went inactive. The churn is accelerating in both directions simultaneously.
Qilin remained the most prolific operator for the second consecutive quarter, posting 414 victims. A group operating under the name “Gentlemen” — which first appeared in leak site data in September 2025, roughly six months earlier — posted 207 victims in Q1 2026, the second-highest total of any group. The group’s targeting spanned financial services firms, hospitals, government agencies and IT providers across Thailand, the United States, France, Brazil and Turkey.
Travelers’ report noted that a fragmented ecosystem is not necessarily a less dangerous one. Disrupting a single dominant group becomes less impactful when attacks are spread across a larger number of operators, each capable of sustaining meaningful volume. The report states that overall ransomware activity has tripled since 2022 based on leak site data.
The AI Governance Gap
While speculation has intensified around AI-enabled cyberattacks — particularly following Anthropic’s announcement of a model the company says outperforms most human researchers at identifying software vulnerabilities — the report argues that the more pressing risk for most organizations is internal. At least 43% of U.S. workers now use AI on the job, a faster adoption rate than personal computers or the internet at comparable points in their commercial lifecycles, the report said.
Organizations are navigating AI deployment through widely varying approaches, Travelers said. Some have imposed restrictions that result in employees using AI tools on personal devices outside company networks, creating data exposure with no visibility into what is being shared. Others have allowed broad, ungoverned adoption, producing unpredictable interactions among overlapping tools.
Even controlled rollouts have produced unintended consequences when employees begin experimenting with AI agents without adequate oversight, which includes agents doing things their creators did not intend, the report said.
The report recommends that organizations establish formal governance before expanding AI use, including at least one individual or committee with accountability for AI decisions, a documented acceptable use policy, mandatory training for employees using generative AI tools and human review requirements for AI-assisted decisions classified as high-risk. The report also recommends integrating AI-specific risk criteria into third-party software procurement and conducting privacy impact assessments on AI tools already in use.
Social Engineering Grows More Sophisticated
Social engineering claims continue to drive significant losses. The severity of claims combining social engineering fraud and business email compromise is up more than 30% since 2023, and those two categories account for roughly 40% to 50% of all cyber claims at Travelers, according to the report.
The tactics themselves are evolving through combination. The report details a pattern it calls “mail bomb + ClickFix,” in which attackers flood a victim’s inbox with thousands of junk messages to create chaos, then contact the victim posing as IT support and walk them through a process that involves pasting a malicious command into the Windows Run dialog, PowerShell or a terminal. The sequence — designed to feel like a help desk interaction — can bypass conventional phishing awareness training because it does not rely on a suspicious link or an urgent password-reset request.
The report identifies procedural controls as the most reliable defense, including verifying inbound IT support contacts through independently sourced channels and treating any instruction to paste commands into a system terminal as a red flag.
Obtain the full report here. &
