Czech Republic - AI Implementation Act

Draft Adaptation Act to Implement the EU Artificial Intelligence Act

Adapční návrh zákona o umělé inteligenci

Czech Republic

RAI-CZ-NA-DAIEAXX-2025
Draft(Being written or scoped)
BillGovernance and OversightConformity Assessment and RegistrationMarket Surveillance
Export PDF

A concise, minimalist draft adaptation law prepared by the Czech Ministry of Industry and Trade (MPO) to implement the EU Artificial Intelligence Act (Regulation (EU) 2024/1689). The draft defines national institutional roles (notifying authority, market surveillance bodies), creates a national regulatory sandbox and sets procedural, language and reporting details needed for domestic enforcement while largely relying on the EU Regulation for substantive rules.

Overview

The Draft Adaptation Act is a short, targeted implementing law prepared by the Czech Ministry of Industry and Trade (MPO) to operationalise Regulation (EU) 2024/1689 (the EU Artificial Intelligence Act) within the Czech Republic. The draft does not replicate the AI Act’s substantive obligations; instead it assigns national institutions, prescribes procedural and language requirements, and establishes enforcement and sandboxing arrangements necessary for the AI Act to function domestically. MPO’s announcement summarizes the approach and rationale; see the Ministry statement at MPO: MPO prepared draft law on AI (26 Sep 2025) and the official VeKLEP/ODok dossier containing the draft and attachments at ODok: Návrh zákona o umělé inteligenci (VeKLEP).

Definitions

The draft uses the definitions contained in Regulation (EU) 2024/1689 for terms such as "AI system", "provider", "operator", "user", "deploy", "placing on the market", "high-risk AI system", and "general purpose AI". It clarifies that where the EU Regulation provides standardised definitions, the national act will apply them verbatim; where the AI Act requires Member State choices (e.g., designation of notifying authorities, language of documentation), the draft supplies those national specifications. The draft also defines national administrative terms (e.g., "oznamující orgán"/notifying authority, "oznámený subjekt"/notified body, "regulační sandbox") in line with the EU text and explanatory materials in the VeKLEP package.

Governance and Institutional Framework

The draft assigns roles and responsibilities across existing Czech agencies rather than creating a new single AI authority. Key appointments and functions in the draft (as explained in MPO and ÚNMZ materials) include: (1) Ministry of Industry and Trade (MPO) as the national coordinator and proposer of the law and national AI policy; (2) Český telekomunikační úřad (ČTÚ) designated as the central market surveillance authority and the public contact point for consumer and market complaints; (3) Úřad pro technickou normalizaci, metrologii a státní zkušebnictví (ÚNMZ) as the notifying authority responsible for designating and supervising notified conformity assessment bodies and registering them with the Commission; (4) Úřad pro ochranu osobních údajů (ÚOOÚ) to exercise market surveillance in areas implicating privacy and data protection; (5) Česká národní banka (ČNB) to supervise AI in entities under its remit (financial institutions); (6) Česká agentura pro standardizaci (ČAS) to establish and operate the national regulatory sandbox. The draft also assigns a human-rights monitoring function to the Office of the Public Defender of Rights (Ombudsman). For MPO’s public announcement see MPO press release and for ÚNMZ’s role see ÚNMZ: Government approved key document (28 May 2025).

Key Focus Areas

The draft focuses on several national choices where the EU AI Act allows discretion: (1) designation of national notifying and market surveillance authorities and the allocation of sectoral surveillance responsibilities (telecoms, privacy, finance); (2) national processes for registration, notification and language of conformity documentation (the draft requires essential conformity documentation and declarations to be available in Czech, Slovak or English when filed with domestic authorities); (3) procedural rules for inspections, requests for information and handling of complaints; (4) creation and governance of a national regulatory sandbox to support innovation while enabling supervised testing and data exchanges with authorities; (5) definitions and administrative categorisation for infractions, including a tailored system for admonitions and reduced fines for minor breaches to support SMEs; (6) accreditation and oversight for notified conformity assessment bodies (to be administered by ÚNMZ and reported to the Commission); and (7) rules on post-market monitoring, mandatory incident reporting and remedies. The draft references EU-level risk classifications (prohibited uses, high-risk systems, etc.)—see the AI Act text at the EU Publications Office for the full regulatory framework (EU AI Act (Regulation 2024/1689) — Publications Office).

Implementation Framework

Implementation is intended through administrative action: MPO will shepherd the bill through inter-ministerial consultation and to government; relevant authorities (ČTÚ, ÚNMZ, ÚOOÚ, ČNB, ČAS) will adopt secondary implementing rules, guidance and technical procedures (e.g., for sandbox admission, notified-body accreditation, and market surveillance coordination). The ODok material package contains the proposed legislative text, the explanatory memorandum and the reasoned justification that show the draft’s minimalist drafting strategy and identify the specific items left for implementing regulation and administrative guidance; access the VeKLEP package at ODok material. The draft also proposes a coordination mechanism for inter-agency cooperation, data-sharing rules for enforcement consistent with data protection law, and a sequence for registering notified bodies with the Commission to enable EU-wide recognition of conformity assessments.

Monitoring and Evaluation

Monitoring combines EU and national mechanisms: providers will continue to file required reports and entries to the EU AI systems database (EU-level), while national authorities will conduct post-market monitoring, inspections and audits. The draft requires domestic authorities to share monitoring results and enforcement data through inter-agency arrangements and to report periodically to MPO. The draft contemplates KPIs for sandbox outcomes, conformity assessment throughput, and enforcement caseloads; it also anticipates a review clause that ties national implementing instruments to EU-level changes (standards, delegated acts and implementing acts).

Penalties, Liability, and Appeals

The draft aligns its administrative offences and sanctions architecture with the AI Act’s enforcement logic but leaves national discretion on the specific administrative penalties in certain cases. It establishes a tiered approach: warnings and remedial orders for minor breaches, administrative fines and suspension measures for serious infringements, and publication of enforcement decisions in more serious or systemic cases. MPO’s public messaging emphasises a proportionate approach with possibilities for admonition and reduced fines for less serious violations to avoid stifling SMEs; see MPO press release. Administrative decisions will be subject to judicial review under Czech administrative procedure rules and applicable appeal routes are provided in the draft explanatory memorandum hosted on ODok.

Relationship to Other Instruments

The draft amends national market surveillance law (it proposes a change to Act No. 87/2023 Sb. on market surveillance) to integrate AI-specific surveillance powers and to allocate responsibilities between existing bodies. It also cross-references Czech data-protection law and the GDPR enforcement roles of ÚOOÚ, banking supervision rules for the ČNB, and sectoral safety laws that interact with AI applications (health, transport). The ODok dossier lists the proposed consequential amendments and difference tables in the material package available at ODok: material.

International Alignment

The draft is explicitly an adaptation of the EU Regulation and is intended to preserve single-market harmonisation while enabling national operational details. Where the AI Act enables Member State choices, the Czech draft selects institutions and administrative procedures so that the Czech approach is interoperable with EU enforcement, notified-body registration and the Commission’s central databases. The draft also preserves compliance with cross-border recognition of conformity assessments and the EU-level prohibition and high-risk framework; it references the EU AI Act text and EU guidance documents available from the Publications Office and Commission pages (EU AI Act — Publications Office).

Implementation Timeline

MilestoneDate / Status
Government decision to transfer coordination to MPO28 May 2025 — government implementation plan approved; see ÚNMZ announcement
MPO public announcement of draft26 Sep 2025 — MPO press release published; see MPO press release
VeKLEP/ODok dossier authorised for circulation25 Sep 2025 — VeKLEP/ODok material created and attachments published; see ODok material
Inter-ministerial consultation closed24 Oct 2025 — status updated on ODok: consultation phase ended
Targeted parliamentary submissionPlanned after inter-ministerial feedback and final MPO revision (anticipated 2026 but not fixed)

Compliance Checklist

ActorKey Immediate Actions
Providers of high-risk AIPrepare conformity documentation in Czech/Slovak/English; arrange conformity assessments; plan registration steps per Annex VIII format; implement risk-management and post-market monitoring systems
Deployers / OperatorsEnsure technical documentation is available; plan human oversight measures; ensure cybersecurity measures; prepare incident reporting workflows
Notified bodies / laboratoriesPrepare accreditation applications to ÚNMZ; align procedures with EU conformity assessment modules
SMEs / StartupsConsider participation in national regulatory sandbox administered by ČAS; use mentoring and sandbox guidance to reduce compliance costs

Sources and References

SourceType
ODok (VeKLEP): Návrh zákona o umělé inteligenci — material package (MPO)Primary Source
MPO press release (26 Sep 2025)Primary Source
ÚNMZ announcement (28 May 2025)Primary Source
EU AI Act (Regulation (EU) 2024/1689) — Publications OfficePrimary Source
Plain English

This draft Czech law sets up the national framework for enforcing the EU Artificial Intelligence Act, affecting anyone developing, deploying, or operating AI systems in the Czech Republic.

The law applies broadly to providers, operators, and users of AI systems, particularly those dealing with "high-risk" AI or "general purpose AI" as defined by the overarching EU regulation. While the EU AI Act lays out the core rules, this Czech adaptation specifies how those rules will be applied locally. Key obligations for businesses include ensuring that essential conformity documentation and declarations are available in Czech, Slovak, or English when submitted to domestic authorities. Providers of high-risk AI systems must also prepare for conformity assessments by nationally designated bodies and establish systems for mandatory incident reporting and ongoing post-market monitoring. Companies should be ready to cooperate with various national market surveillance authorities, which are assigned specific oversight roles.

The law is currently a draft, with an anticipated effective date of January 1, 2026. Enforcement will involve a tiered approach, starting with warnings and remedial orders for minor breaches, escalating to administrative fines and suspension measures for serious infringements. The Ministry of Industry and Trade (MPO) has indicated a proportionate approach, including possibilities for admonition and reduced fines for less serious violations, aiming to support small and medium-sized enterprises (SMEs).

A practical consideration for businesses is the fragmented enforcement landscape. Instead of a single AI authority, oversight is distributed across several existing agencies: - The Czech Telecommunication Office (ČTÚ) will act as the central market surveillance body. - The Office for Personal Data Protection (ÚOOÚ) will handle privacy-related AI issues. - The Czech National Bank (ČNB) will supervise AI in financial institutions. - The Office for Technical Standardization, Metrology and State Testing (ÚNMZ) will designate and oversee conformity assessment bodies. This multi-agency approach means businesses might need to interact with several different regulators depending on their AI system's application, which could add complexity to compliance and appeals processes. The draft also establishes a national regulatory sandbox, run by the Czech Standardization Agency (ČAS), offering a supervised environment for testing innovative AI systems.

Plain-English rewrite by Regulations.ai — not legal advice. Verify against the official text.

What you must do — compliance checklist

0 / 12 marked complete

Plain-English obligations under Czech Republic - AI Implementation Act. Not legal advice — verify against the official text before relying on it.

  1. #1CriticalBy 2026-01-01

    Applies to: Providers of high-risk AI systems.

    Prepare conformity documentation in Czech/Slovak/English
  2. #2CriticalBefore placing on market

    Applies to: Providers of high-risk AI systems.

    arrange conformity assessments
  3. #3CriticalBefore placing on market

    Applies to: Providers of high-risk AI systems.

    plan registration steps per Annex VIII format
  4. #4CriticalBefore placing on market

    Applies to: Providers of high-risk AI systems.

    implement risk-management and post-market monitoring systems
  5. #5CriticalBefore placing on market

    Applies to: Providers of high-risk AI systems.

    implement risk-management and post-market monitoring systems
  6. #6CriticalBy 2026-01-01

    Applies to: Notified bodies and laboratories.

    Prepare accreditation applications to ÚNMZ
  7. #7CriticalBy 2026-01-01

    Applies to: Notified bodies and laboratories.

    align procedures with EU conformity assessment modules
  8. #8ImportantBefore deploying AI system

    Applies to: Deployers and operators of AI systems.

    Ensure technical documentation is available
  9. #9ImportantBefore deploying AI system

    Applies to: Deployers and operators of AI systems.

    plan human oversight measures
  10. #10ImportantBefore deploying AI system

    Applies to: Deployers and operators of AI systems.

    ensure cybersecurity measures
  11. #11ImportantBefore deploying AI system

    Applies to: Deployers and operators of AI systems.

    prepare incident reporting workflows
  12. #12Recommended

    Applies to: SMEs and startups developing AI.

    Consider participation in national regulatory sandbox administered by ČAS

© Regulations.AI — created on 13-Jun-2026