We use cookies on this page. View our Privacy Policy for more information.

NIS2 and the Czech Cybersecurity Act: what they mean for AI-assisted development

NIS2 and the Czech Cybersecurity Act: who may be covered, which deadlines run and what it means for companies developing with AI. For CEOs and CTOs.

NIS2 and the Czech Cybersecurity Act: what they mean for AI-assisted development

Short answer: NIS2 is the EU directive on cybersecurity, and in the Czech Republic it is implemented by the Cybersecurity Act, in force since 1 November 2025. It mainly concerns mid-sized and larger companies in regulated sectors and, indirectly, their suppliers. A company checks for itself whether it is covered, registers with NÚKIB (the national cybersecurity authority) and, within a year of the registration decision, introduces measures under the higher or lower obligations regime. For a company that develops with AI, this in practice means knowing what is running, who has access to what, what is logged and how code is checked before deployment. Check the specific impact on your company with a lawyer.

Why it matters

For part of the market, cybersecurity has stopped being voluntary. The NIS2 directive, formally Directive (EU) 2022/2555, covers 18 sectors according to the European Commission and makes top management accountable for failing to follow risk-management measures. The Czech Republic implemented it through a new Cybersecurity Act, in force since 1 November 2025. NÚKIB expected roughly 6,000 organisations to fall under the act, and as of 3 June 2026, 5,768 of them were on record.

In 2026 the act is no longer just on paper. NÚKIB sends registration decisions to companies that have registered, and from the moment of delivery a one-year deadline runs for introducing the measures and for incident reporting. The authority has also started carrying out inspections: according to its reply from June 2026 it opened seven inspections this year, and its 2026 inspection plan covers 33 entities.

For a CEO this has two consequences. If your company is covered, the deadlines are running. And even if you are not, your larger customers may be, and they will ask similar things of you as a supplier. In both cases the security of company development, including development with AI, is a topic for management. This article is an overview of principles, not a legal analysis. We verified the situation on 4 November 2026.

Not sure whether the act applies to you, or what it means for your AI projects? Book 30 minutes online. We will go through what runs in your company and where the biggest risks are. Book an online call

Who may fall under NIS2 and the act

What decides it is the service you provide and how large you are, not whether you feel like "critical infrastructure". According to the European Commission, NIS2 covers 18 sectors: besides energy, transport, health, banking, water and digital infrastructure also, for example, digital service providers, IT service management, waste management, manufacturing, postal services, public administration or the space sector. The obligations apply mostly to medium-sized and large entities.

The Czech act translates this into the concept of a regulated service and into two regimes: higher and lower obligations. Which regime applies depends on the type of service and the size of the company. A company verifies on its own whether it is covered, and NÚKIB offers a calculator on its portal for that. The deadline for registration ended on 31 December 2025. According to NÚKIB, whoever missed it should register as soon as possible, because the length of non-compliance affects any penalty.

For a software company there is another way in: the supply chain. NIS2 counts on managing the security of suppliers. If you build software for a bank, a hospital or an energy company, you will probably meet part of the requirements by contract even if you are not covered by the act yourself.

What obligations the act imposes in general

The obligations can be summed up in four areas. They are similar in both regimes and differ in depth. The specific measures are set by implementing decrees (separately for the higher and lower regime), and their scope for your company is a question for a lawyer.

  • Risk management. A company must know what assets it has (systems, data, applications), what the risks are and what measures it has put in place. As an ongoing process, not a one-off document.
  • Supplier security. Cloud services, external developers and AI services are part of your risk. A company must know who has access to what, and have it covered in contracts as well.
  • Incident reporting. Incidents are reported through the NÚKIB Portal. According to a summary of the act, the first report is submitted without undue delay and at the latest within 24 hours of discovery, followed by a notification within 72 hours and a final report within 30 days. The scope of reporting differs by regime. Without logs and monitoring you will not meet such deadlines, because you will not even recognise the incident.
  • Management accountability. NIS2 introduces accountability of top management. Management is expected to approve the measures, oversee them and understand them. Security therefore cannot simply be handed to IT.

Penalties come on top of that. For serious violations NÚKIB states a fine of up to CZK 250 million or 2% of net worldwide annual turnover. So far the authority has punished sparingly, says its aim is to help, and by June 2026 had by its own account imposed one fine. That is still no reason to postpone the deadlines.

What it means for development with AI

The act does not ban development with AI and does not care what you write the code with. But it makes visible things that easily stay in a grey zone in a company where people develop with Claude Code, Cursor or Lovable. The four obligations above take a concrete form in such a company.

  • Inventory. Risk management starts with knowing what runs in the company. An application that a department built itself and that runs on someone else's hosting with company data is an asset management does not know about, and therefore a risk it does not manage. We call this shadow AI and cover it in the article Shadow AI.
  • Architecture and permissions. A bigger risk than the code itself is often what an application or an AI agent has access to. An agent with direct write access to a production database or a key to a payment system is a problem even if its code is clean. Access and keys belong in central management, not in code or on a laptop, and each system should have only the permissions it really needs. We write more about this in the article AI agents for development in your company, and about injected instructions in Prompt injection and AI agents.
  • Logging. Reporting within 24 hours assumes you record the incident at all. Applications written with AI often have no logging or only basic logging, as we discuss in the article on the security of apps written with AI. Without logs you cannot detect, report or investigate an incident.
  • Checks before deployment. AI can write secure code if it gets the right brief and rules, but it makes mistakes much like a person. In October 2026 Veracode tested 18 models: they wrote secure code in 62% of cases on average, the best models (Claude Opus 5.5 and GPT-6-Sol) in 73%. A sensible approach is automated AI code review of every change first, and then a person who focuses on the substantial risks.

The level of checking is determined not by whether the application goes to customers, but by the impact of an error and whether it can be undone. A wrong write to a database you can restore from a backup; money sent or data leaked you cannot take back. Testing is set accordingly: a penetration test makes sense when ordered according to risk, that is, according to the type of application, its architecture, the sensitivity of data and the impact of a possible incident. For a web application it then focuses mainly on the backend, the API, login and access rights. When and how to order a pentest is covered in the article Penetration test of an app built with AI.

At MeguMethod we handle this within AI Ops: we map what runs in the company, set the guardrails and oversee the code.

How NIS2 relates to the AI Act and GDPR

The three regulations overlap, and it pays to handle them together. NIS2 and the Cybersecurity Act deal with the resilience of systems and incident reporting. The AI Act deals with what an AI system may do and how transparent it must be. GDPR deals with the personal data that systems process, including how to delete or hand them over securely, which is mainly a question of the backend and a clear data-management procedure. One incident, for example a data leak from a company AI agent, can trigger obligations under several regulations at once. We will cover the AI Act and GDPR in separate articles.

An inventory of systems, access management and logging help with all three. The rules also keep evolving: in January 2026 the European Commission proposed targeted changes to NIS2 meant to simplify compliance, and as of the date of this article the proposal is in the legislative process. Check the specific impact on your company with a lawyer.

When NIS2 is your topic and when it is not

You can watch it from a distance for now if:

  • you operate outside regulated sectors and have no customers who fall into them,
  • you are a small company below the size thresholds and do not supply regulated services,
  • the NÚKIB calculator came out negative for you and you have it documented,
  • you do not develop software that runs at regulated customers.

Deal with it now if:

  • you operate in one of the 18 NIS2 sectors and are a medium-sized or larger company,
  • the calculator came out positive and you have not registered yet,
  • you have a registration decision and your one-year deadline for measures is running,
  • you supply software or services to companies that fall under the act,
  • your company develops with AI, nobody has an overview of what runs, and AI tools have wide access to data.
If you recognised yourself in the second list, send us a few sentences about what runs with AI in your company. We will get back to you with a proposal on how to proceed. Send an inquiry

How MeguMethod approaches it

We do not provide legal assessment of whether and in which regime a company is covered; that belongs to a lawyer. We do the technical part. We use AI from the product and technical specification onwards, and every project has rules for AI: architecture, limits and restrictions. We work with AI across the whole development; only the level of checking changes with the risk.

Within AI Ops we proceed in four steps. We map what runs in the company, who builds it, where it is hosted and what it touches. We set guardrails: stack, environments and deployment rules. We equip and train people. And we keep oversight of the code, where every change is checked by AI and by engineers for security, architecture and functionality.

This addresses the risks that overlap with the act: leaks of sensitive data, dependence on systems of a single author, duplicate development and applications that cannot be maintained. It includes controlled deployment only into an approved company environment, central management of access and keys, and closed AI loops. If a company needs an application built and operated, we offer App as a Service, and for AI assistants over company data AI assistants. We start with a 30-minute consultation whose output is a list of risks and next steps.

Frequently asked questions

Does NIS2 apply to a small software company too?

Directly, usually not, as the obligations mostly concern medium-sized and large entities in regulated sectors. Indirectly yes, if you supply software to regulated customers who must manage supplier security. Check whether you are covered with the NÚKIB calculator and with a lawyer.

Since when does the act apply and which deadlines run?

The act came into force on 1 November 2025. Registration of a regulated service was due by 31 December 2025. From the delivery of the registration decision, according to a summary of the act, a company has one year to introduce security measures and start incident reporting.

What happens if a company does not register?

NÚKIB states that entities that did not register face an administrative penalty with a fine of up to CZK 250 million or 2% of net worldwide annual turnover, and that the length of non-compliance affects the size of the penalty. At the same time it offers help and guidance on how to register retroactively.

Do we have to stop using AI development tools because of NIS2?

No. The act does not deal with what you write code with, but with whether you know what runs, who has access to it, whether anyone has checked it and whether you can recognise an incident. With rules, limited permissions and checks according to risk, development with AI is compatible with the act.

Do we need a penetration test because of the act?

Not automatically. A pentest makes sense according to risk: the type of application, its architecture, the sensitivity of data and the impact of an incident. Whether and how often to test in your regime, check with a lawyer or a security specialist.

Next step

Want to keep development with AI under control with NIS2 in mind? See how AI Ops works, or book 30 minutes right away and we will go through what runs in your company. If you want the context first, read how to get vibe coding in a company under control.

We will go through what runs with AI in your company and where the risks are. Book an online call

Sources

  • European Commission: NIS2 Directive, page updated 2 July 2026. digital-strategy.ec.europa.eu
  • NÚKIB: Ohlášení podle nového zákona o kybernetické bezpečnosti provedlo přes 4 800 organizací (registrations under the new Cybersecurity Act exceeded 4,800 organisations), 11 February 2026. nukib.gov.cz
  • NÚKIB: Informace ke kontrolní a sankční praxi NÚKIB (reply under the Freedom of Information Act), 3 June 2026. nukib.gov.cz
  • NÚKIB: Informace k regulaci a řízení kybernetické bezpečnosti (reply under the Freedom of Information Act), 23 June 2026. nukib.gov.cz
  • Právní prostor: Nový zákon o kybernetické bezpečnosti: Klíčové změny pro firmy (key changes for companies), 24 September 2025. pravniprostor.cz
  • Lupa.cz: NÚKIBu se ani po uplynutí lhůty nenahlásily všechny regulované firmy (not all regulated companies registered even after the deadline), 11 February 2026. lupa.cz
  • European Parliament, Legislative Observatory: procedure 2026/0012(COD), proposed changes to NIS2. oeil.europarl.europa.eu
  • Veracode: Veracode Finds First Meaningful Gain in AI Code Security Since 2025 as Average Pass Rate Hits 62%, 29 October 2026. veracode.com
  • MeguMethod: AI Ops, service page. megumethod.com/ai-ops

FAQ

Does NIS2 apply to a small software company too?

Directly, usually not, as the obligations mostly concern medium-sized and large entities in regulated sectors. Indirectly yes, if you supply software to regulated customers who must manage supplier security. Check whether you are covered with the NÚKIB calculator and with a lawyer.

Since when does the act apply and which deadlines run?

The act came into force on 1 November 2025. Registration of a regulated service was due by 31 December 2025. From the delivery of the registration decision, according to a summary of the act, a company has one year to introduce security measures and start incident reporting.

What happens if a company does not register?

NÚKIB states that entities that did not register face an administrative penalty with a fine of up to CZK 250 million or 2% of net worldwide annual turnover, and that the length of non-compliance affects the size of the penalty. At the same time it offers help and guidance on how to register retroactively.

Do we have to stop using AI development tools because of NIS2?

No. The act does not deal with what you write code with, but with whether you know what runs, who has access to it, whether anyone has checked it and whether you can recognise an incident. With rules, limited permissions and checks according to risk, development with AI is compatible with the act.

Do we need a penetration test because of the act?

Not automatically. A pentest makes sense according to risk: the type of application, its architecture, the sensitivity of data and the impact of an incident. Whether and how often to test in your regime, check with a lawyer or a security specialist.

AppCast

Czech apps conquering the world

Recipes for successful apps, in Czech. MeguMethod CEO Vratislav Zima talks business with the founders of Czech apps.

Gradient noisy shapeMegumethod CEO Vratislav ZimaMegumethod Scribble