Resources, Security
NIS2 and AI agents: what the directive means for your AI use
NIS2 does not mention AI agents, but if your company is in scope, its agents and AI vendors fall under the same cybersecurity duties as any other system or supplier. This guide explains the duties and how they apply to AI.
Updated 20 November 2026. 9 minute read.
In short
- NIS2, Directive (EU) 2022/2555, covers medium-sized and large entities in the sectors listed in its Annexes I and II, plus some entities of any size. They are either essential or important entities.
- Article 21 requires appropriate and proportionate risk-management measures, including incident handling, supply-chain security, access control, encryption and multi-factor authentication.
- Significant incidents must be reported with an early warning within 24 hours, an incident notification within 72 hours and a final report within one month.
- Management bodies must approve the measures, oversee them and follow training, and they can be held liable. AI agents are network and information systems, and AI vendors are suppliers, so both fall within these duties.
- Denmark implemented NIS2 through Act No 434 of 6 May 2025, in force since 1 July 2025.
Who is in scope: essential and important entities
The NIS2 Directive, Directive (EU) 2022/2555, replaced the first NIS Directive. Member States had to transpose it by 17 October 2024 and apply it from 18 October 2024. Because it is a directive, the duties reach companies through national laws.
As a rule, NIS2 covers entities in the listed sectors that are at least medium-sized: 50 or more employees, or an annual turnover and balance sheet total both above €10 million. Some entities are covered whatever their size, such as qualified trust service providers, top-level domain name registries and DNS service providers.
- Annex I, sectors of high criticality: energy, transport, banking, financial market infrastructures, health, drinking water, waste water, digital infrastructure, ICT service management (business to business), public administration and space.
- Annex II, other critical sectors: postal and courier services, waste management, chemicals, food, manufacturing of certain products, digital providers such as online marketplaces, search engines and social networks, and research.
Large entities in Annex I sectors are generally essential entities. Most other entities in scope are important entities. Both must meet the same risk-management and reporting duties. The difference lies in supervision and fines: essential entities face proactive supervision and maximum fines of at least €10 million or 2% of worldwide annual turnover, whichever is higher. Important entities are supervised after the event, with maximum fines of at least €7 million or 1.4%.
The risk-management measures in Article 21
Article 21(1) requires appropriate and proportionate technical, operational and organisational measures to manage the risks to network and information systems and to prevent or minimise the impact of incidents. The measures must follow an all-hazards approach. Article 21(2) lists at least ten:
- policies on risk analysis and information system security;
- incident handling;
- business continuity, such as backup management, disaster recovery and crisis management;
- supply-chain security, including security aspects of relationships with direct suppliers and service providers;
- security in acquiring, developing and maintaining systems, including vulnerability handling and disclosure;
- policies and procedures to assess the effectiveness of the measures;
- basic cyber hygiene practices and cybersecurity training;
- policies on cryptography and, where appropriate, encryption;
- human resources security, access control policies and asset management;
- multi-factor or continuous authentication, secured voice, video and text communications, and secured emergency communications where appropriate.
For digital infrastructure, ICT service and digital providers, such as cloud providers, data centres and managed service providers, Commission Implementing Regulation (EU) 2024/2690 sets detailed technical requirements. ENISA published non-binding technical implementation guidance for it in June 2025, with a mapping to standards such as ISO/IEC 27002. Other entities can also use it as a reference.
Reporting: 24 hours, 72 hours and one month
Article 23 requires entities to notify their CSIRT or competent authority of any significant incident. An incident is significant if it has caused or can cause severe operational disruption or financial loss, or if it has affected or can affect others by causing considerable material or non-material damage.
- Early warning: without undue delay and within 24 hours of becoming aware of the incident, saying whether it is suspected to be caused by unlawful or malicious acts or could have a cross-border impact.
- Incident notification: within 72 hours of becoming aware, updating the early warning with an initial assessment of severity and impact and, where available, indicators of compromise.
- Intermediate report: on request of the CSIRT or authority.
- Final report: no later than one month after the incident notification, with a detailed description, the type of threat or root cause, the mitigation measures and any cross-border impact. If the incident is still ongoing, a progress report is due instead, with the final report within one month of handling it.
Where appropriate, entities must also inform the recipients of their services without undue delay about significant incidents that are likely to affect them.
Management accountability and supply-chain security
Management bodies
Article 20 makes cybersecurity a board-level duty. Management bodies must approve the risk-management measures, oversee their implementation and can be held liable for breaches. Their members must follow training, and entities are encouraged to offer similar training to employees regularly. For essential entities, competent authorities can, as a last resort, seek a temporary ban on managers at chief executive or legal representative level from managerial functions if breaches are not remedied (Article 32(5)).
Supply-chain security
Under Article 21(3), entities must take into account the vulnerabilities specific to each direct supplier and service provider, the overall quality of their products and cybersecurity practices, and their secure development procedures. Article 22 also allows Union-level coordinated risk assessments of critical ICT supply chains.
Supplier security is not a one-off questionnaire. Assess suppliers before you sign, write security duties into contracts, and review them when the service or the risk changes.
Where AI agents and AI vendors fit
NIS2 is technology-neutral. An AI agent that reads email, updates records or calls internal systems is part of your network and information systems. The model provider and the agent platform are direct suppliers or service providers. So the Article 21 duties apply to them in the same way as to any other system or vendor.
- Risk analysis: include each agent in your risk assessment, with the data and systems it can reach. Agent-specific risks, such as prompt injection and excessive agency, belong here.
- Access control: give each agent only the access its role needs, with its own credentials. Protect the accounts that manage agents with multi-factor authentication.
- Cryptography: check that the credentials agents use are stored encrypted.
- Supply chain: assess model providers and agent platforms on security practices, data location, sub-processors and incident notice. Ask how fast they will tell you about an incident, given your own 24-hour clock.
- Incident handling: an agent that leaks data or makes unauthorised changes may cause a significant incident. Your plan should cover pausing agents, revoking credentials and preserving logs.
- Business continuity: decide what happens to the work if a model provider is unavailable.
- Training: cover AI agents in cyber hygiene training for employees, and in the management body's training.
- Effectiveness: review agent permissions and logs on a schedule, and test agents against misuse.
If you also deploy high-risk AI systems, the EU AI Act adds its own cybersecurity and logging requirements. The two regimes overlap but do not replace each other.
Denmark's implementation
Denmark implemented NIS2 through the Act on measures to ensure a high level of cybersecurity (NIS 2-loven), Act No 434 of 6 May 2025. It entered into force on 1 July 2025, after the EU deadline of 17 October 2024. Separate acts cover the energy and telecommunications sectors. Financial entities covered by the Digital Operational Resilience Act (DORA) follow its sector-specific rules.
- Covered entities had to register on Virk.dk by 1 October 2025. Entities that missed the date should still register.
- The Ministry of Resilience and Preparedness coordinates implementation. Supervision is split between sector authorities, including the Danish Agency for Digital Government.
- The Agency for Digital Government publishes guidance in English on who is covered, the requirements and incident notification.
Some details, such as which authority supervises a given sector and later executive orders, continue to develop. Check the Danish authorities' official guidance and the act on Retsinformation for your sector before relying on a specific point.
A NIS2 checklist for AI agents
- Confirm whether your company is an essential or important entity in each country where it operates, and register where required.
- Add every AI agent to your asset inventory and risk assessment, with its owner, data and access.
- Add model providers and agent platforms to your supplier register and assess them.
- Write incident notice, security duties and audit rights into AI vendor contracts.
- Give each agent its own least-privilege credentials, stored encrypted.
- Require multi-factor authentication for employees who manage agents.
- Log agent actions and keep the logs for incident investigation.
- Update the incident plan so the 24-hour early warning is possible for agent-caused incidents.
- Plan for a model provider outage.
- Brief the management body on AI risks and record its approval of the measures.
- Review permissions, suppliers and logs on a fixed schedule.
How Blentera helps
Blentera is a platform where companies run AI agents with employees in charge. Each agent has a role in the org chart and can use only the apps and tools it is given. Approval rules decide which actions need an employee's approval, and budgets warn at 80% and pause the agent at the limit. Credentials are encrypted with AES-256-GCM. Employees sign in with passkeys or two-step verification, with single sign-on on Business and user provisioning (SCIM) on Enterprise. The activity log can be exported for investigations, and the platform is hosted in the EU. These controls support your NIS2 work. Meeting NIS2 remains your company's responsibility.
Questions
- NIS2 does not regulate AI as such. If your company is in scope, its AI systems are part of its network and information systems, and its AI vendors are suppliers, so the Article 21 risk-management and supply-chain duties apply to them.
- An early warning within 24 hours of becoming aware of a significant incident, an incident notification within 72 hours, and a final report no later than one month after the incident notification.
- Both must meet the same risk-management and reporting duties. Essential entities face proactive supervision and higher maximum fines, of at least €10 million or 2% of worldwide turnover. Important entities are supervised after the event, with maximum fines of at least €7 million or 1.4%.
- Article 20 says management bodies must approve and oversee the cybersecurity measures and can be held liable for breaches. National laws set out how. For essential entities, managers can also be temporarily banned from managerial functions in some cases.
- Yes. Denmark's NIS 2 act, Act No 434 of 6 May 2025, entered into force on 1 July 2025. Covered entities had to register on Virk.dk by 1 October 2025.
Sources
- Directive (EU) 2022/2555 (NIS2 Directive), EUR-Lex
- Commission Implementing Regulation (EU) 2024/2690, EUR-Lex
- NIS2 Directive, European Commission
- NIS2 technical implementation guidance, ENISA, June 2025
- NIS2 Directive: Denmark, European Commission
- Lov om foranstaltninger til sikring af et højt cybersikkerhedsniveau (NIS 2-loven), lov nr. 434 af 6. maj 2025, Retsinformation
- What is NIS 2?, Danish Agency for Digital Government
- Who is covered by the NIS 2 law?, Danish Agency for Digital Government
- Registration obligation, Danish Agency for Digital Government
- Incident notifications, Danish Agency for Digital Government
Related
This guide is general information, not legal advice. Check how the rules apply to your company with your own advisers.
Get started
Put your first agent to work in minutes.
Three steps, and your first agent is working for your company.
Step 1
Create an account
Sign up with your work email and invite your team. No sales call needed.
Step 2
Add your first agent
Create an agent, give it a role, and connect your own AI key and the apps it may use.
Step 3
Give it a task
Ask in Blentera or your chat app, and follow the work until it is done.