NIS2 and AI: what the directive asks of your use of AI
Short answer
NIS2 does not regulate AI, but it does touch it: an AI service is part of your network and information systems and of your supply chain. Essential and important entities must manage the risks of their AI suppliers, give an early warning of significant incidents within 24 hours, and their management body is accountable for those measures. The Netherlands and Germany have transposed NIS2; France had not by September 2026.
Where NIS2 stands in 2026
The NIS2 Directive (EU) 2022/2555 had to be transposed by October 2024. Most member states were late. The situation in three countries: The reporting deadlines and supplier terms are in NIS2 incident reporting and AI suppliers.
| Country | Status |
|---|---|
| Netherlands | The Cybersecurity Act (Cyberbeveiligingswet) has applied since 15 August 2026, together with the Critical Entities Resilience Act |
| Germany | The NIS2 Implementation Act (NIS2UmsuCG) has applied since 6 December 2025, without transition periods; registration with the BSI |
| France | Not yet transposed in September 2026; the European Commission referred France to the Court of Justice on 8 July 2026. ANSSI published a voluntary reference framework (ReCyF) in March 2026 |
If your organisation falls within scope, you must register, take appropriate security measures and report significant incidents.
Where NIS2 and AI meet
- Risk management (Article 21): AI applications belong in your risk analysis, like any other information system. Which data goes in, who has access, what if the service fails?
- Supply chain security: an outside AI service is a supplier. NIS2 asks you to assess suppliers’ security practices and agree on them. With a large foreign AI vendor you have little leverage there.
- Incident reporting (Article 23): a significant incident requires an early warning within 24 hours, a notification with an initial assessment within 72 hours and a final report within a month. An AI incident, such as data leaking through an AI service, can qualify.
- Management body (Article 20): it approves the measures, oversees their implementation, must follow training itself and can be held liable.
Five questions your board must be able to answer about AI
- Which AI services does the organisation use, officially and unofficially? See Shadow AI.
- Which data goes into them, and with which suppliers does it sit?
- What happens if an AI service fails or changes its terms?
- Can we detect and report an incident with an AI service within 24 hours?
- Can we show who did what with AI? See The EU AI Act in 2026 for the overlap with the AI Act.
Shortening the supply chain
The more outside AI services, the more suppliers you must assess, contract and monitor. Running AI on your own infrastructure removes that link from the chain: there is no outside model vendor that processes data or can fail.
With werqly, open-weight models run on a node of your own in your server room, linked to your identity provider. The werqly AI Control audit log goes to your SIEM, so an AI incident falls under the same detection and reporting procedure as the rest of your systems. Self-managed, werqly is mainly a software supplier; with werqly Pro management or werqly Cloud, werqly is a supplier in your chain that you assess like any other. werqly’s ISO 27001 certification is in progress. See also What is private AI?
Frequently asked questions
Questions on this topic
Does NIS2 apply to AI?
Not as AI regulation, but AI services are part of the network and information systems NIS2 covers. Organisations in scope must include AI in their risk management, assess AI suppliers and report significant incidents involving them.
Is an AI service part of the NIS2 supply chain?
Yes. An outside AI service processes your data and is part of your information systems. NIS2 asks you to assess the security of such suppliers and agree on it, as with other ICT suppliers.
Must a data leak through an AI chatbot be reported under NIS2?
If it is a significant incident, yes: an early warning within 24 hours and a notification with an initial assessment within 72 hours. Separately, the GDPR breach notification duty to the data protection authority applies.
Is the board personally accountable under NIS2?
The management body must approve the security measures, oversee their implementation and follow training itself. It can be held liable if it fails to do so. AI risks therefore belong on the board agenda.
Sources
- NIS2 Directive — Directive (EU) 2022/2555
- Dutch NCSC — Cyberbeveiligingswet (NIS2), Dutch
- BSI — NIS-2-Umsetzungsgesetz in force (German)
- ANSSI — MonEspaceNIS2 (French)
This article is general information, not legal advice. Rules and guidance change; check the sources listed or ask your lawyer or DPO if in doubt.