2026 DIVD — agentic AI intrusion via Zammad zero-days (CVE-2026-102489/102490); data exfil
Data compromised
Scope unpublished. DIVD states it cannot rule anything out and is treating the incident as worst-case assume-breach until proven otherwise. Directly involved parties informed; reported to Autoriteit Persoonsgegevens and Dutch NCSC; options discussed with police. Individual census unpublished.
Technical writeup
Verified DIVD CSIRT notice — September 24, 2026 (“It was a matter of when, not if…”). Dutch Institute for Vulnerability Disclosure detected suspicious activity, concluded it had been compromised, blocked infrastructure access, and engaged a third-party IR team. DIVD describes the modus operandi as an agentic AI-powered attack not previously seen in its experience; investigation ongoing; worst-case assume-breach posture. Directly involved parties informed; incident reported to Autoriteit Persoonsgegevens and the National Cyber Security Centre; discussed with police. Next public update scheduled Monday September 28 (or sooner). Spokesperson Marieke Rijken (communications@divd.nl). Headcount and data inventory unpublished — recordsAffected 0; companyConfirmed true. Follow-up (BleepingComputer Sep 30, 2026): DIVD attributed initial access to Zammad flaws CVE-2026-102489 and CVE-2026-102490 enabling session hijacking, remote code execution, and root escalation within seconds via agentic automation; data was read/exfiltrated; deeper lateral movement limited by segmentation. DIVD/Merlon notified Zammad; users urged to upgrade to Zammad 7 or take instances offline.
Root cause
Chain of two Zammad ticketing zero-days (session hijack + RCE/privilege escalation) exploited by an autonomous AI agent
References
- https://csirt.divd.nl/2026/09/24/when-not-if/
- https://databreaches.net/2026/09/25/divd-dutch-institute-for-vulnerability-disclosure-investigating-agentic-ai-powered-attack/
- https://www.bleepingcomputer.com/news/security/divd-says-zammad-zero-days-enabled-ai-driven-network-breach/
- https://www.bleepingcomputer.com/news/security/automated-ai-agent-used-to-breach-cybersecurity-nonprofit-divd/