DIVD publishes the two Zammad zero-days used in the breach of its own systems that it blames on AI agents

The Dutch Institute for Vulnerability Disclosure (DIVD), which warns organisations about exposed and vulnerable systems, published case DIVD-2026-00015 on 30 September: two vulnerabilities in Zammad, the open-source helpdesk software, found while it investigated a breach of its own systems. The case page says Zammad was abused to breach DIVD on 21 September.

DIVD's site lists that breach as a separate case under the line “DIVD got hacked through AI agents”, and BleepingComputer reported on 30 September that DIVD says the Zammad zero-days enabled an AI-driven network breach. The new case page itself does not describe the agents' role.

CVE-2026-102489 is a session hijack that leads to remote code execution as the zammad user. It is exploitable in Zammad 6.3.0 to 6.5.4, and present but not exploitable in 7.0.0 to 7.1.3. CVE-2026-102490 lets the local zammad user escalate privileges to root, in versions from 1.5.0 to 7.1.0-alpha. BleepingComputer reports DIVD saying that together they let attackers hijack sessions, run code remotely and get to root in seconds.

DIVD's timeline: analysed and reproduced on 22 and 23 September, reported to Zammad on 24 September, and on 26 September it began scanning for exposed instances, made a limited disclosure and started notifying owners. Its advice: “We advise all users of Zammad to upgrade to version 7 of Zammad or to take it offline.” The case gives no count of vulnerable servers, and no Zammad advisory for these CVEs was visible from here.

• Confirmed CVE-2026-102489 (session hijack to remote code execution, exploitable in 6.3.0–6.5.4) and CVE-2026-102490 (local escalation to root, 1.5.0–7.1.0-alpha). DIVD
• Confirmed Zammad was abused to breach DIVD on 21 September; reported to Zammad on 24 September; owners notified from 26 September. DIVD
• Reported DIVD says the Zammad zero-days enabled an AI-driven breach of its network. BleepingComputer