Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

AI risk watchHappening now

An agentic AI attack breached a vulnerability disclosure nonprofit in seconds

ResearchersAi enabled attacks

The Dutch Institute for Vulnerability Disclosure (DIVD) says attackers chained two Zammad zero-days, CVE-2026-102489 and CVE-2026-102490 (both CVSS 9.4), to get remote code execution and root on its help desk on September 21, then pivoted and took data. DIVD describes the modus operandi as an agentic AI-powered attack. Segmentation limited how far it got.

Why it matters

The victim is a security organization, the entry point an everyday support tool, and the chaining took seconds. Patch windows built around human attackers are now too long.

What you can do

  • Keep internet-facing helpdesk and ticketing systems off the core network and patch them first.
  • Segment so one compromised service cannot reach the rest; DIVD's segmentation is what limited this.

Sources

See it with the other risks in the Learning Center