Why Can Your Printer Talk to Your Domain Controller?
Find Threat Talks on
No patch, no defense. That’s the usual thinking about zero-days. So what decided how far an AI agent got inside the Dutch Institute for Vulnerability Disclosure (DIVD), the people who warn the internet about its vulnerabilities?
The agent came in through two zero-days in Zammad, the DIVD’s ticketing system, and tried to go further. Some data got out. The rest of the network didn’t, because the ticketing system was never allowed to reach it.
Rob Maas, Field CTO at ON2IT, talks with Gabriel Reis, SOC analyst at ON2IT and DIVD volunteer, about the first three days, why the DIVD went public before it knew the full story, and the questions your organization can start asking on Monday.
What you’ll learn
- You can’t patch a zero-day, but you decide what it reaches. At the DIVD, segmentation limited how far the attacker got after the ticketing system fell.
- This AI attack was noisy. The agent left notes explaining its actions, and one engineer answered with prompt injection planted on his own servers.
- Silence buys attackers time. Early disclosure lets customers watch for phishing built on their stolen data.
Your cybersecurity experts
Gabriel Reis
Junior SOC Analyst at ON2IT
Rob Maas
Field CTO at ON2IT
Episode details
If your defense depends on knowing the vulnerability first, a zero-day wins by definition. The DIVD had two of them in Zammad, its ticketing system: nothing to patch, no signature to match. What limited the damage was a design choice made long before the attack. The ticketing system could not freely talk to everything else, so the attacker hit walls before it reached the core.
Gabriel’s question cuts to it: why is the printer allowed to talk to the domain controller? Can you answer that for your own “boring” systems? They still hold data an attacker wants. For the DIVD, that meant knowledge of open vulnerabilities and who still needs to fix them, what Rob calls toxic data and an attacker would call gold.
The second lesson is about the clock. The DIVD went public within three days and assumed everything was compromised until it knew more. That is assume breach applied to communication. Companies often hold back for legal reasons and to protect credibility, but every day of silence gives attackers time to use stolen contact data for phishing while customers have no reason to be suspicious. The reaction to the DIVD’s statement was positive.
The AI angle cuts both ways. The attack was run by an AI agent, messy enough to leave notes behind, and one engineer answered with prompt injection aimed at attacking agents. Machine versus machine speeds up the fight, which makes Gabriel’s closing questions more urgent: which systems do you have, how do they communicate, and what could an attacker reach from inside them tomorrow? That is where Zero Trust starts.
Get your Hacker T-shirt
Join the treasure hunt!
Find the code within this episode and receive your own hacker t-shirt for free.
