How does a NIS2 risk assessment with AI work?

What does the AI assistant do in the risk assessment?

An AI assistant connects to your company in the app through MCP (Model Context Protocol), an open standard. It fills in the NIS 2 walkthrough as the person who connected it, with that person's rights.

In 2.1, “Decide how you rate risks”, it can adopt the ready-made method from BSI Standard 200-3, just as you would with “Adopt”. The method has four levels for frequency, four for damage and four risk levels.

In 2.3, “Rate your risks”, it sets two values for each entry: how often the worst realistic case happens and how big the damage would be. If asked, it adds a note, for example what you already do about it.

The matrix turns that into a level: low, medium, high or very high. Until you have checked them and management signs off, these are proposals.

Art. 21(2)(a) NIS 2 · § 30(2) no. 1 BSIG

How does the assistant know your company's risks?

An assistant knows what your records say. That the inventory system went down twice last month, or that a supplier would be hard to replace, it knows only if someone wrote it down.

You bring that judgement. Think of fire, water and power cuts, and remember that rare does not mean small: in the BSI matrix, damage that threatens the company's existence is never low.

How do you check the assistant's proposals?

Check every row on the three screens “Rate your software”, “Rate your technology and rooms” and “Rate your suppliers”. Every screen an assistant wrote shows its name and the day.

The overview “Your risks at a glance” then shows how many entries sit in each field of the matrix. Below it, all of them are listed by name, highest level first.

You change a rating there by tapping the name or the field.

What does management do in item 7.3?

In 7.3, “Record the management review”, management sees the risk map on the screen “Your risks for management”: where your software, technology and suppliers stand today. It decides where to act first and records in “Your management review” what has to be done, by whom and by when.

Under “What the answers still leave open”, the walkthrough shows what is still missing by your own answers. Only fixed cases appear there by themselves, for example a supplier rated high or very high whose contract settles neither security nor incidents. Another is a backup with no restore that worked.

For each open point, you set in the app who takes care of it and by when. For its digital services, the Implementing Regulation requires identifying this for the risk treatment measures (Annex point 2.1.2(i)).

No assistant fills this part: the management review, the open points and the sign-off are done by a person in the app. Signing off is for whoever holds the management role in the app, signed in with their own account. Why the law leaves this decision with management is explained in our article on who answers for AI entries.

Art. 20(1) NIS 2 · § 38(1) BSIG

What happens when something changes after the sign-off?

If a person changes a rating in the app after the sign-off, item 2.3 goes back to management, because the sign-off was given on the old values. For a supplier this applies to 5.1 and 5.2 as well. Only sign-offs older than the change are taken back.

While an item that relies on the ratings or the supplier list is signed off, such as 2.3, 5.1 or 5.2, the tools refuse the assistant's rating. The refusal names the items.

The same goes for the method in 2.1, because adopting it replaces the scales the ratings sit on. Taking a sign-off back is a person's step: you make the change in the app, and the item goes to management again.