- Threat actors exploited vulnerabilities in 28% of ransomware cases, up from 14% in Q2
- Patch volumes increased almost 80% as vendors remediate weaknesses identified by AI
- Detailed analysis of one intrusion conducted almost entirely by AI agents showcases AI coordination and sophistication
- Healthcare remains the sector most affected by ransomware, with 21% of investigated cases
WEST HARTFORD, Conn., Nov. 19, 2026 /PRNewswire/ -- In its Quarterly Threat Report for Q3 2026, Beazley Security, a company of Zurich, announced that reported ransomware incidents rose 38% from the second quarter and that the share of its ransomware cases that began with an exploited vulnerability doubled to 28%. The shift comes as defenders face soaring software update and vendor patch volumes, with published CVEs increasing almost 80% to nearly 37,000 in the third quarter.
"At today's volume, security teams simply cannot treat every vulnerability as equally urgent," said Alton Kizziah, CEO of Beazley Security. "Start with known exploited vulnerabilities and internet-facing applications, then prioritize systems that provide access to credentials, cloud infrastructure or critical business operations."
Attackers are finding the systems organizations overlook
While patch volumes from major technology companies such as Microsoft and Oracle are soaring, many ransomware incidents that began with exploitation involved self-hosted, internet-facing applications such as build servers, analytics platforms and collaboration tools, which can get less attention than operating systems and endpoint software. In several cases, attackers found cloud access keys on compromised servers and used them to move further in and, in some instances, deploy ransomware through the victim's cloud tools.
Malware accounted for another 19% of initial access, up from 14%, driven primarily by ClickFix, which uses a fake error message or CAPTCHA to persuade users to run attacker-supplied commands. Ransomware or extortion generally followed a ClickFix compromise by two to three weeks, a window in which isolating systems and rotating credentials can stop an attack.
An intrusion conducted almost entirely by AI agents
The report details an intrusion carried out largely by coordinated AI agents. Patching and MFA stopped its first attempts, but the agents got in through a flaw in a custom application and stole data.
Notes the attackers left behind showed a multi-agent harness working across three servers. The agents launched about a dozen known exploits, but the targeted systems had been patched, and MFA stopped their credential attacks. AI agents then exploited a flaw in a custom-built .NET application, found a service account credential and used it to move laterally and steal data.
Beazley Security Labs details the case in "LARA: What an AI Attack Harness Wrote Down," available at https://labs.beazley.security/articles/lara-what-an-ai-attack-harness-wrote-down.
"LARA isn't a one-off. Open-source attack harnesses like ARTEX are already showing up in real intrusions," said Francisco Donoso, Chief Product and Technology Officer at Beazley Security. "These agents use the same old tricks. They just find the unpatched server or the flat network faster than any human team, and they never get tired. That's why patching, segmentation, and basic hygiene matter more than ever."
How attackers got around MFA
Adversary-in-the-Middle kits, which sit between a user and the real sign-in page, captured session cookies after users authenticated. Device-code phishing persuaded victims to approve attacker access on Microsoft's genuine sign-in page. Infostealers took session tokens along with passwords, and attackers sometimes registered their own devices or MFA methods to keep access after resets.
"Attackers have learned to wait until the user signs in, then steal the session," said Josh Carolan, Director of Security Research at Beazley Security. "MFA still matters, but it can't protect a session that's already been stolen."
Beazley Security recommends organizations harden their MFA by implementing passkeys and phishing resistant MFA, disabling device-code authentication where it is not required, restricting new MFA registration methods from unmanaged devices, and requiring compliant, organization-managed devices through conditional access.
About the report
The Quarterly Threat Report combines findings from Beazley Security's incident response investigations, threat intelligence research, and managed detection and response telemetry. It also covers business email compromise, infostealers and emerging ransomware operators. Read the full report at https://beazley.security/insights/quarterly-threat-report-third-quarter-2026.
About Beazley Security
Beazley Security, a company of Zurich, is a global cybersecurity services firm offering managed detection and response, incident response, exposure management and advisory services. Beazley Security helps organizations understand their exposure, defend against threats, respond to incidents and strengthen long-term resilience.
Beazley Security Labs publishes ongoing threat research, vulnerability advisories and technical analysis at labs.beazley.security.
SOURCE Beazley Security

Share this article