Does NIS2 cover employee AI usage?

Yes, indirectly but bindingly. NIS2 Art. 21 requires risk management including supply-chain security, and external LLM providers are part of the ICT supply chain the moment employees or systems send them data. An uncontrolled AI channel is an unmanaged supply-chain risk, and executives carry personal liability for it. NativeAI Guard gives essential entities monitoring and enforcement over that channel.

Is an AI data leak a reportable incident under NIS2?

A significant incident affecting an essential entity's services or data must be reported under NIS2 Art. 23, and AI-channel leaks or compromised agents can qualify. NativeAI Guard contributes detection and evidence: every AI interaction is logged, so an incident can be identified, scoped and documented.

Our Copilot Enterprise is governed. What about the rest of the web?

That is the gap group CISOs name most often: Copilot Enterprise covers the Microsoft estate, while ChatGPT, Claude, DeepL and every web-based AI form sit outside it with no monitoring or enforcement. NativeAI Guard covers this web surface: prompts and uploads are inspected in the browser and on the desktop before data leaves.

Can operational data like grid reports leak through AI tools?

Yes. An engineer uploading an operational report to an LLM for summarization sends network topology, capacity data and incident details to a consumer AI service. NativeAI Guard scans documents inside the upload and blocks infrastructure-sensitive content before it leaves the organization.

Industries / Critical infrastructure · energy, utilities, transport

AI protection for critical infrastructure. Protect your employees' AI usage.

NIS2 and CER make cybersecurity risk management (including the AI supply chain) a legal duty for critical infrastructure entities, with personal liability for executives. Meanwhile, your AI attack surface is entirely uncontrolled.

DEADLINESOCT 2024 · NIS2 transposition passed; enforcement ramping upJUL 2026 · CER critical-entity designationAUG 2026 · EU AI Act (safety components of critical infrastructure)

The AI governance framework for critical infrastructure

Five obligations. Here's our part, honestly.

We mark what NativeAI Guard solves for the channels we protect, where it contributes, and what stays with your processes.

NIS2 Art. 21Risk management incl. supply-chain security, LLM providers are part of your ICT supply chain.✓ We solve · AI traffic control

NIS2 Art. 23Incident reporting, leaks via the AI channel and compromised agents are reportable events.◐ We contribute · detection + evidence

CER DirectiveResilience duties for critical entities, the same monitoring and logging controls, second mandate.◐ We contribute · monitoring + reports

EU AI ActAI in safety components of critical infrastructure is high-risk from Aug 2026.◐ We contribute · logging + oversight

Swiss DSG / GDPREmployee and customer data in AI workflows, minimization, records, transfer limits.✓ We solve · anonymization + RoPA

Seen in energy & industry

The uncontrolled web attack surface

Copilot Enterprise covers the Microsoft environment. ChatGPT, Claude, DeepL and every web form sit outside it, no monitoring, no governance, no enforcement. The gap every group CISO names.

NativeAI Guard: covers everything else, the web and AI attack surface your existing stack cannot inspect.

The uploaded grid report

An engineer uploads an operating report to an LLM to get a summary. Network topology, capacity data, incident details, infrastructure-critical content in a consumer AI tool.

NativeAI Guard: documents are scanned inside the upload, sensitive operational data detected, the file blocked before it leaves the company.

The infected page

A web page ingested by an internal RAG pipeline contains hidden instructions, and the agent obediently starts acting on them. Prompt injection is invisible to every traditional security tool you run.

NativeAI Guard: XPIA blocking strips malicious instructions from everything your agents ingest. NativeAI Guard Agent DLP & Firewall →

The web attack surface is not controlled.

GROUP CISO · EUROPEAN ENERGY GROUP

Policies, managed

The rulebook for critical infrastructure is already written. You adapt it instead of inventing it.

A base rulebook is already written and active from day one: the rules from the Swiss DSG, GDPR and the EU AI Act, plus the requirements your industry is specifically subject to. It covers the standard cases without you writing a single line. Security and compliance practitioners review it, and we keep it current as the law changes. On top of it sit the rules that apply only to you: your own data structures, your internal requirements, your exceptions.

Critical infrastructure packNIS2 Art. 21 · 23CER directiveOT and safety dataSwiss DSG · GDPREU AI Act, safety components

During onboarding we tune the pack to your own data structures and internal rules, working from the compliance documents you already have.

Common questions

Frequently asked questions

Our driver is NIS2 and CER, not Swiss data protection. Does that change anything?

It changes the framing, not the mechanism. NIS2 makes cyber risk management, including your AI supply chain, a legal duty with personal accountability for management, and CER adds resilience duties. What you need is the same: visibility into an uncontrolled channel, enforcement, and an incident record you can report from within the required window. Swiss hosting is often the reason customers come to us in the first place. For NIS2, though, it does not count: the directive does not ask where a system runs, but whether you can manage your risks, detect incidents and report them on time. That is exactly what visibility, enforcement and the incident record deliver, regardless of where the servers stand.

Our exposure is engineers uploading whole technical documents, not typing secrets into a chat box. Is that covered?

Yes. Documents are inspected inside the upload, at the moment of upload, before they leave the company: an operating report with network topology, capacity data and incident details is caught as a file, not only as text someone might paste. Password-protected files are intercepted rather than waved through. Document upload was named unprompted by a group CISO in the energy sector as the thing that worried him most.

How do we demonstrate traceability and run a data protection impact assessment when nobody knows what is being sent?

That is the reporting layer rather than the enforcement layer. Every interaction produces a record of who sent what to which model and what was done about it, which is the raw material a data protection impact assessment and an AI inventory both require. Organizations in this sector typically tell us they have no formal AI governance framework yet; the passive pilot phase exists precisely to produce that baseline before protection is switched on.

Protect your employees' AI usage, with a pilot phase that we run for you.

A report for management on data leaks, shadow AI, AI usage and model costs, across the whole web attack surface: the evidence your NIS2 risk assessment is missing.