NIS2:resilience,incidentsandthesupplychain
For essential and important entities, an AI system is another dependency in scope. The relevant questions are about resilience, reporting and the chain behind the platform.
ControlsrelevanttothemeasuresNIS2addresses
Risk analysis and policies
Risk classification per workflow, with policy expressed as enforced configuration rather than documentation.
Incident handling
Detection signals on agent behaviour, tested kill switches, and evidence records usable as investigation material.
Business continuity
Defined degradation behaviour per workload class, and tested recovery paths.
Supply-chain security
Provenance for container images, model artefacts, dependencies and MCP servers, with version pinning.
Network security
Default-deny egress, segmentation and the asymmetric IT/OT boundary.
Access control
Per-agent non-human identity, short-lived credentials and default-deny data scope.
Cryptography
Encryption in transit and at rest, with customer key custody available.
Vulnerability handling
Disclosure and patch processes, and the update path for disconnected deployments.
Effectiveness assessment
Evidence of denied attempts and evaluation results as measurable input to control review.
Training
Operator and approver documentation, because oversight roles that nobody understands are not controls.
NeuroCluster supports controls relevant to this framework. Whether a particular deployment meets a particular obligation depends on how it is configured, operated and documented, so a deployment-specific assessment remains required. Nothing on this page is legal advice.
Questions
- Would you be in scope as a supplier to us?
- That depends on your sector, your entity classification and how the platform is deployed. It is a determination for you, and we will provide what your supply-chain assessment requires.
- How do you support incident notification timelines?
- Through detection signals, evidence available immediately rather than after reconstruction, and contractual notification terms agreed per deployment. The obligation and the clock remain yours.
- Does an AI system increase our attack surface?
- Yes, and pretending otherwise would be unhelpful. An autonomous caller with system access is a new class of principal. The controls on the security page exist because that is true, and the honest framing is risk managed rather than risk absent.
Bring us one operational problem.
You do not need a finished brief. Bring the problem — we will work out the next step together.
Or book a call with the team