NIS-2 Directive
An EU cybersecurity law that requires a broad range of "essential" and "important" organizations to manage risk across their supply chain — including the third-party vendors and AI tools they send data to — or face fines that scale with global turnover.
What Is the NIS-2 Directive?
The NIS-2 Directive (Directive (EU) 2022/2555) is the European Union's updated network and information security law, replacing the original 2016 NIS Directive with a substantially wider scope and stricter obligations. Where the original directive covered roughly 10,000 entities, NIS-2 is estimated to bring more than 160,000 entities across the EU into scope, extending well beyond traditional critical infrastructure sectors like energy and transport into areas such as healthcare, digital infrastructure, food production, postal services, and manufacturing.
The directive splits covered organizations into two tiers — essential and important entities — based on sector and size, with essential entities typically having 250 or more employees or exceeding €50 million in annual turnover, though certain entities such as trust service providers and critical infrastructure operators are covered regardless of size. Both tiers face largely the same underlying obligations, including a set of minimum cybersecurity measures under Article 21 covering technical, operational, and organisational risk management, incident reporting within tight timelines, and — notably for organizations that rely on external AI vendors — an explicit requirement to manage supply chain and third-party security risk, not just risk inside the organization's own systems.
Practical Industrial Use
A hospital network classified as an essential entity under NIS-2 is a clear example of where the directive's supply chain requirements intersect directly with AI vendor use. If the hospital sends patient data to an external AI tool for clinical documentation or research, NIS-2's requirement to assess and manage risk across its supply chain extends to that vendor relationship — meaning the hospital needs to be able to account for what data reaches the vendor and how it's protected, not just how its own internal systems are secured.
The same considerations apply broadly across NIS-2's expanded scope: a manufacturing company classified as an important entity evaluating an AI tool for production data analysis, a financial market infrastructure provider assessing AI-assisted document review against its supply chain risk obligations, or a digital infrastructure provider documenting how sensitive operational data is protected before it reaches any third-party AI system. In each case, an organization's NIS-2 obligations don't stop at its own network boundary — they extend to the vendors, including AI vendors, that its data flows through.
What Happens Without It
Organizations in scope of NIS-2 that adopt AI tools without accounting for supply chain risk are exposed to a distinct kind of compliance gap: the directive doesn't just require securing an organization's own systems, it requires demonstrating that risk introduced through third-party relationships — including AI vendors — has been assessed and managed. Failing to do so isn't a hypothetical concern; NIS-2 penalties can reach up to €10 million or 2% of global annual turnover for essential entities, whichever is higher, and enforcement has been active, with the European Commission pursuing member states over transposition delays and pressing entities toward compliance as the directive's incident-reporting and registration deadlines have taken effect.
⚠ Risk Without NIS-2 Compliance This becomes a particularly acute gap when sensitive data — patient records, financial data, operational details tied to critical infrastructure — is sent to an AI vendor without first assessing what that vendor receives and how it's protected, since an incident or exposure at the vendor can trigger the same reporting and accountability obligations as an incident within the organization's own systems, while top management is held directly accountable for the gap either way.
With Supply Chain Risk Addressed
- Data sent to AI vendors is assessed and protected as part of the organization's broader NIS-2 supply chain risk management obligations
- Sensitive identifiers can be masked or protected before reaching a vendor, reducing the risk surface the vendor relationship introduces
- Organizations can document what data reaches third-party AI tools and how it's protected, supporting the evidence NIS-2 compliance requires
- Incident risk tied to vendor relationships is reduced at the source, rather than discovered only after an incident occurs
Without It
- AI vendor relationships become an unmanaged part of the organization's supply chain risk, in tension with NIS-2's explicit supply chain obligations
- An incident or exposure at the vendor can trigger the same reporting and liability exposure as an internal incident
- Fines that scale with global turnover apply regardless of whether the failure originated inside the organization or through an unmanaged third party
- Top management accountability under NIS-2 extends to vendor relationships that were never properly assessed
How This Relates to Questa AI
Questa AI helps organizations in scope of NIS-2 manage exactly the kind of third-party risk the directive's supply chain requirements are aimed at: by detecting and masking sensitive identifiers before data reaches an external AI vendor, Questa reduces what any given AI vendor relationship actually exposes the organization to. This is closely related to Questa's support for local and self-hosted deployment, since organizations classified as essential or important entities can keep the anonymization process itself within infrastructure they directly control, further limiting the supply chain risk introduced by adopting AI tools.
This approach is particularly relevant for organizations that need to document — not just assert — how they've addressed the risk an AI vendor relationship introduces, since Questa's Blackbox recording provides a record of what was detected and protected before transmission, and the governance dashboard offers visibility into where in the pipeline that protection is applied. Together, these give NIS-2-covered organizations a way to build evidence supporting their supply chain risk management obligations, rather than relying solely on a vendor's own assurances.
Frequently asked questions
NIS-2 applies to medium and large entities operating in specific sectors, split into essential entities (high-criticality sectors like energy, transport, banking, and healthcare) and important entities (a broader set including food production, postal services, and manufacturing), with some entities covered regardless of size.
NIS-2's Article 21 risk management measures explicitly extend to supply chain security, meaning covered entities need to assess and manage the risk introduced by vendors — including AI vendors — that process their data, not only risk within their own systems.
Penalties for essential entities can reach up to €10 million or 2% of global annual turnover, whichever is higher, with important entities facing somewhat lower but still substantial fines, alongside other enforcement measures such as warnings and binding instructions.
The transposition deadline was 17 October 2024, but implementation across member states has proceeded unevenly, with the European Commission pursuing infringement proceedings against some member states for delays, meaning specific obligations can vary somewhat depending on the member state.
Not automatically, but it creates an obligation to assess and manage that relationship as part of the organization's supply chain risk — sending sensitive data to a vendor without that assessment is where the compliance gap arises, not the use of an AI tool itself.
No single measure satisfies NIS-2 in isolation. Protecting sensitive data before transmission is one part of managing supply chain risk under Article 21, alongside broader requirements like incident reporting, risk assessments, and governance accountability.
Related terms
Local Redaction
Removing or masking sensitive data on the device or within the organization's own environment before anything is ever transmitted to an external AI model — protection that happens before the data leaves, rather than trusting a third party to handle it responsibly once it arrives.
Third-Party Data Exposure
The risk that sensitive or regulated data is disclosed to, or accessed by, an external vendor, partner, or AI provider beyond what the originating organization intended or authorized — often as a byproduct of routine data sharing rather than a security breach.
Controlled Cloud Environment
A cloud infrastructure setup where an organization — not a third-party AI vendor — dictates exactly where data is processed, how long it's retained, who can access it, and which regulatory boundaries it never crosses, turning data residency and access control from a vendor's policy into the organization's own enforceable configuration.
Zero Data Exposure
"Zero" is doing a lot of work in that phrase — and whether it's backed by real architecture or just confident marketing copy is exactly what a buyer needs to verify before trusting it.
Cyber-Sensitive Data
The category of information that isn't sensitive because it identifies a person or a business secret, but because it maps out how to break in — credentials, network architecture, vulnerability details, and security configurations that turn an AI tool's normal output into an attacker's shortcut if handled carelessly.
See NIS-2 Directive in practice
Questa AI anonymizes sensitive data before it reaches any AI model — across documents and live prompts, with governance and data-residency control.