tlder@dev — An AI agent chained two Zammad zero-days to break into a disclosure group
tlder@dev:~$
Cross-Cutting/security, ai-ml

An AI agent chained two Zammad zero-days to break into a disclosure group

  • Discussion

The chain is nasty, and now it's on CISA's exploited list. Session fixation (CVE-2026-102489) hands an attacker code execution as the helpdesk's service user; a second bug (CVE-2026-102490) walks that up to root. CISA added both on October 2, after DIVD found its own ticketing instance breached — not by a person at a keyboard, but by an autonomous AI agent that ran the whole intrusion itself. The maintainers hardened the code in version 7.2.0, which is the fix. If you self-host this helpdesk, upgrade now and treat anything older as suspect — DIVD describes data theft and root access, not a lab proof-of-concept. The agent angle is what makes this one linger. It's an early, concrete case of a model driving a real attack end to end, not just flagging the bug for a human to exploit. Patch, rotate sessions, and hunt for the root foothold.