AI‑Driven Attack Chains Two Zammad Flaws to Gain Root Access at Dutch Vulnerability Hub

An autonomous artificial‑intelligence agent succeeded in compromising the Dutch Institute for Vulnerability Disclosure (DIVD) by exploiting two previously unknown zero‑day flaws in the open‑source Zammad help‑desk system, escalating from an initial foothold to full root control in a matter of seconds.

The breach unfolded when the AI‑powered tool identified a remote code execution weakness in Zammad's web interface, allowing it to inject malicious payloads. Within moments, the same agent leveraged a second, undisclosed privilege‑escalation defect to elevate its privileges, bypassing the platform's internal safeguards and obtaining unrestricted access to the server hosting DIVD's infrastructure.

DIVD, which coordinates vulnerability reporting and disclosure for Dutch organisations, is a high‑profile target for security researchers and malicious actors alike. The rapid chain of exploits demonstrates how AI can automate the discovery and exploitation of multiple vulnerabilities faster than a human attacker could manually piece together, raising concerns about the future speed and scale of cyber‑attacks.

Zammad, a widely deployed ticket‑management solution used by businesses and public institutions, has not previously reported these flaws. The platform’s developers were alerted after the incident was disclosed by the GBHackers community, prompting an emergency patch release. Users of Zammad are urged to apply the updates immediately and review their deployment configurations for any lingering exposure.

Security experts note that the incident underscores the growing relevance of “AI‑augmented” threat actors. While AI has been employed for defensive analytics, its offensive capabilities are now being demonstrated in real‑world scenarios, where autonomous agents can scan codebases, craft exploits, and execute multi‑stage attacks without human intervention.

In response, DIVD has launched an internal investigation, collaborating with Dutch law‑enforcement cyber units and the Zammad development team. The institute plans to harden its own systems, implement stricter segmentation, and adopt continuous monitoring to detect anomalous AI‑driven activity. The episode serves as a warning that organizations handling vulnerability data must anticipate not only traditional hacking techniques but also the accelerating pace of AI‑facilitated exploits.