At the ONE Conference in The Hague, the new NCSC director warned about autonomous AI agents that scan networks independently and launch targeted phishing attacks. He explicitly called IT service providers and MSPs gatekeepers for the Netherlands' digital safety. What that means if you've outsourced your IT.
At the ONE Conference in The Hague, in early October, the new director of the Dutch National Cyber Security Centre gave his first major public appearance. Matthijs van Amelsfort, in post since 1 September and previously with the national police, did not strike a reassuring tone. He warned about what he called unfounded naivety regarding cyber risks, combined with the rapid rise of autonomous AI agents. Together, he said, these developments are a worrying signal for companies and institutions.
What stands out is not just the warning itself, but who he holds partly responsible for acting on it. Not only organisations themselves, but explicitly their IT service providers and MSPs too. He called them gatekeepers for the Netherlands' digital safety. For a director who has outsourced IT, that is not an abstract statement from The Hague. It touches directly on the question of who answers for what once something goes wrong.
The threat Van Amelsfort describes is concrete. Where an attacker used to manually explore a network, hunt for a weak spot, and work through trial and error to get in, an autonomous AI agent does this on its own. Such an agent scans networks continuously, writes its own scripts to exploit a vulnerability, decides from that which attack route offers the best odds, and then launches a phishing attack tailored to the specific target.
The difference with automated attacks from a few years ago lies in that independence. An older style of attack script follows a fixed sequence of steps and stumbles as soon as something deviates from the pattern. An AI agent adapts its approach to what it encounters along the way, much as a human attacker would, except around the clock and at almost no cost. For organisations that rely on human response time, for example an administrator reviewing logs each morning, that shifts the window in which an incident needs to be caught.
Van Amelsfort called cyber resilience a team sport, where defence does not rest only with technical specialists but also with management and other stakeholders inside an organisation. Investments in security only work when they have broad backing, not when they land on the desk of one systems administrator. In the same message, he pointed to the role of IT service providers and MSPs: for many Dutch organisations, they manage the infrastructure an attack hits first, which means their choices help determine how resilient the whole picture is.
That is not an optional role. An IT partner who rolls out patches on time, keeps access tightly managed, and recognises unusual behaviour on a network before it causes damage does not just keep its own client standing. It also helps keep up a piece of the wider digital infrastructure that other organisations, suppliers, and ultimately government depend on.
The statement doesn't stand on its own. Since 15 August, the Cyberbeveiligingswet, the Dutch implementation of NIS2, has applied without a transition period to thousands of organisations providing essential or important services. Part of those obligations also reaches organisations that don't fall directly under the law, through the procurement terms their clients apply. An IT service provider supplying a client that does fall under the law often sees those same requirements passed straight through into the contract.
The NCSC director's message adds a practical layer that reaches further than the eight thousand organisations directly covered by the law. Anyone outside that list still depends on an IT partner that takes its gatekeeper role seriously, simply because an attack doesn't respect the boundaries of legislation.
A gatekeeper is only useful if it actually guards the gate. Four questions help test that with your own IT partner, whether it sits in-house or is outsourced. How quickly are security updates rolled out across servers, workstations, and network equipment once a vendor releases them, and is there an agreed timeframe for that.
Is the network continuously watched for unusual behaviour, or only after something has already gone wrong. Is there a concrete plan for the moment an incident occurs, including who you call and what happens in the first hours. And finally, how is access to systems limited to what someone actually needs for their work, so that one stolen password doesn't open the door to everything.
If you get concrete, explainable answers to these questions, you have a partner that fills its gatekeeper role. If the answers stay at reassuring generalities, that is exactly the kind of unfounded naivety Van Amelsfort warned about, now shifted from your own organisation to the party you hired to prevent it.
You don't need to understand every technical detail of autonomous AI agents yourself to act on this. What does help is not letting the conversation with your IT partner end at the annual invoice, but periodically checking whether monitoring, patch policy, and incident response have kept pace with a threat that has itself become faster and more independent. A gatekeeper that was sufficient two years ago doesn't have to be sufficient now.
Zarioh manages exactly the infrastructure, monitoring, and incident response this gatekeeper role requires for its clients. Want to have your current IT environment checked against it, or considering placing this role with a dedicated partner? Feel free to get in touch for a no-obligation conversation.
Zarioh Digital Solutions
IT specialists from Utrecht, the Netherlands. We help businesses with Microsoft 365, AI agents, hosting and telephony, and share what we learn in practice. Follow us on LinkedIn

Regelgeving & Compliance

Regelgeving & Compliance

Regelgeving & Compliance