AI-driven hackers exploit Zammad flaws to steal researchers’ contact details
The Dutch Institute for Vulnerability Disclosure (DIVD), an organisation that hunts security bugs professionally, was hacked via two zero-day vulnerabilities in its Zammad support platform software. The attackers used the flaws to hijack sessions, execute remote code and escalate privileges to root access within seconds, stealing email addresses and contact details of DIVD's volunteer security researchers. This matters because DIVD identified the operation as AI-powered or agentic, indicating malicious actors are using autonomous AI agents to carry out complex, multi-stage cyberattacks.
The two vulnerabilities, CVE-2026-102489 and CVE-2026-102490, carry CVSS severity scores of 9.4 when chained together. The attack occurred on 21 September and was discovered the following day. DIVD staff noted unusual patterns—rapid-fire automated actions, sloppy logic and embedded notes in the attack script explaining the attacker's own reasoning—that indicated AI involvement rather than human hacking. Zammad versions 6.3.0 through 6.5.4 and 7.0.0 through 7.1.3 are vulnerable to the first flaw; all Zammad versions are vulnerable to the second. The organisation has advised all users to upgrade to version 7 or take Zammad offline.
- AI agents hacked a security research organisation via two critical Zammad flaws.
- Volunteer researchers' contact details stolen; increased risk of social engineering.
- Attack was unusually loud and messy, with AI leaving notes justifying its actions.
New here? Start with this
The Dutch Institute for Vulnerability Disclosure is an organisation that searches for security bugs in software—before criminals can find them first. Last month, the institute's own computer systems were hacked through two previously unknown flaws in Zammad, a support software it uses to manage enquiries from clients and researchers. The attackers stole email addresses and contact details belonging to the institute's volunteer security researchers.
The most significant aspect of this attack is that the hackers appear to have used artificial intelligence to carry it out. Instead of a human hacker, an AI system automatically identified the software flaws, broke through the system's defences and extracted the data in seconds. The institute identified this as AI-driven because the attack showed distinctive signs: rapid automated actions, errors in the logic and notes written into the code that appeared to explain the AI's own reasoning.
The problem extends far beyond the Dutch Institute because Zammad is used by organisations around the world. Zammad has advised users to upgrade to the latest version of their software or take it offline completely. This incident demonstrates that criminals are increasingly using AI to carry out cyber-attacks automatically.
Read the full article at the source →
Originally published by The Register as “AI agents hacked the hackers, stealing email addresses from security research org”.