The breach at the Dutch Institute for Vulnerability Disclosure (DIVD)
Late last week the Dutch Institute for Vulnerability Disclosure (DIVD) announced it had been hacked after seven years of largely uneventful operations. DIVD is a volunteer-run nonprofit that scans the internet for systems affected by known vulnerabilities, notifies the systems' owners, and provides mitigation guidance. The organization described the incident as serious, and said evidence left by the attacker had been sufficient to help reconstruct parts of the intrusion.
An automated, agentic AI attacker
DIVD’s investigators described the intrusion as “an agentic AI-powered attack.” According to the organization, the attacker used an automated AI agent that acted autonomously on DIVD’s network: “after every action it decided the next step itself, at the speed of light and sloppy logic or pattern.” The group added the agent “did some pretty dumb things,” such as interfering with its own adversary-in-the-middle activity via password spraying and over-explaining its decisions in embedded comments.
Audit-ready is a season. It shouldn't be.
Evidence in spreadsheets, controls drifting between audits, frameworks multiplying on flat headcount. Nubivance runs continuous compliance on Rapid7 Cyber GRC - SOC 2, HIPAA, ISO 27001, PCI, CMMC.
End the scrambleThe exploited "technical vulnerability" — explicitly not Citrix NetScaler
DIVD said the attacker first exploited a “technical vulnerability” in an undisclosed system, then used the AI agent to carry out post-exploitation activity. DIVD explicitly stated the flaw leveraged in the intrusion was not Citrix NetScaler. The organization withheld the precise nature and patch state of the vulnerability while investigators continue their work, and said it will notify other possible victims of the same vulnerability “as soon as they can.”
Investigation and official notifications: police, Autoriteit Persoonsgegevens, and the NCSC
DIVD has opened an internal investigation and informed law enforcement and regulatory bodies. The organization notified the police, the Autoriteit Persoonsgegevens (the Netherlands’ data protection authority), and the National Cyber Security Center (NCSC). DIVD said it would provide a more detailed update on October 1. BleepingComputer contacted DIVD seeking further details about the vulnerability and its patch state but had not heard back as of publication.
What this means for technologists, regulators, and affected system owners
- Technologists and security teams: DIVD says the agent left ample forensic evidence because it was poorly trained and configured, making reverse-engineering of its actions possible. That residual evidence could aid technical teams trying to map the attacker’s steps and to determine which systems or data, if any, were affected.
- Regulators and law enforcement: The Autoriteit Persoonsgegevens, the NCSC, and the police have been notified. Those bodies will receive DIVD’s findings as the organization completes its investigation and prepares the promised October 1 update.
- Affected enterprises and system owners: DIVD has pledged to notify other possible victims of the same vulnerability “as soon as they can,” signalling that organizations running similarly configured systems should be alert for outreach from DIVD and for any advisories that emerge from the October 1 update.
The attack marks a conspicuous example of a security researcher organization itself being targeted by what it calls an agentic AI. DIVD’s description — an autonomous agent that chose steps in real time, made clumsy operational choices, and left verbose commentary — leaves two concrete next steps on the calendar: the group’s promised and more detailed update on October 1, and its notification process to inform other potentially affected parties of the undisclosed vulnerability.
Read the original report on BleepingComputer: https://www.bleepingcomputer.com/news/security/automated-ai-agent-used-to-breach-cybersecurity-nonprofit-divd/
