Hungary - EU AI Regulation Implementation (LXXV/2025)

Overview

Act LXXV of 2025 establishes Hungary's domestic legal and institutional framework to implement Regulation (EU) 2024/1689 (the EU AI Act). The Act was promulgated in Magyar Közlöny No. 127 (2025. Oct. 31) and creates the national roles and mechanisms required by the EU AI Act: the national notifying authority for conformity assessment, the AI market surveillance authority that will perform single point of contact and enforcement tasks, and an advisory Hungarian Artificial Intelligence Council. The Act is available in the parliamentary publication: text of the Act (PDF) and the Országgyűlés legislative dossier provides the parliamentary metadata and PDF link. The Act is designed to align Hungarian enforcement and institutional arrangements with EU obligations while allowing national procedural rules and sectoral delegation (e.g., the Magyar Nemzeti Bank for financial sector high‑risk systems).

Definitions

The Act adopts and cross‑refers to the definitions of the EU AI Act for key concepts ("AI system", "provider", "deployer", "placing on the market", "putting into service", "high‑risk AI system", "general‑purpose AI model (GPAI)", "regulatory sandbox", etc.). It explicitly clarifies that "use of an AI system output in Hungary" may bring the matter within Hungarian jurisdiction even where the system was developed or placed on the market outside Hungary. The Act also defines domestic institutional terms ("AI market surveillance authority", "notifying authority", "Hungarian Artificial Intelligence Council", "regulatory sandbox operator").

Governance and Institutional Framework

The law designates the National Accreditation Authority (NAH) as the national notifying authority under Article 28 of the EU AI Act and designates a ministerial office as the AI market surveillance authority (press coverage and the implementing decree identify the relevant ministry). The market surveillance authority will operate the national single point of contact, supervise post‑market surveillance, operate the regulatory sandbox and coordinate with sectoral surveillance authorities. For sectoral, prudential or specialist domains (e.g., finance), the Magyar Nemzeti Bank (MNB) retains market surveillance competence where the AI system is used in the direct provision of financial services. The Act also establishes the Hungarian Artificial Intelligence Council as an advisory, cross‑sectoral body chaired by the government AI official; the Council has powers to issue guidance and to request data from authorities for regulatory purposes. See the parliamentary PDF and analyses summarising the institutional design: Act text (PDF) and commentary summarised by legal advisors.

Key Focus Areas

The Act implements the EU risk‑based approach: (1) prohibits unacceptable AI practices defined by the EU AI Act; (2) sets out obligations for high‑risk systems (conformity assessment, documentation, technical risk mitigation, human oversight, logging and record‑keeping, cybersecurity, data governance and registration); (3) imposes transparency obligations on certain limited‑risk and generative or general‑purpose models (notably disclosures when interacting with an AI or using AI‑generated material); and (4) exempts minimal‑risk applications from special restrictions but encourages voluntary good practice. The Act emphasises market surveillance, conformity assessment infrastructure (through the NAH and notified conformity assessment bodies), and regulatory sandboxes/regulatory test environments to support innovation under supervision. National measures clarify jurisdictional reach and the treatment of downstream deployers and wrappers. Several press and professional summaries explain the scope and obligations in practical terms; examples and sectoral guidance are expected from the Hungarian Artificial Intelligence Council and from national authorities.

Implementation Framework

Operational responsibilities and many procedural rules are delegated to an implementing government decree (the Act anticipates designation of authorities, the single point of contact operation, and detailed procedural rules for market surveillance and administrative sanctions). The NAH will run the system for designating and supervising conformity‑assessment bodies; the market surveillance authority will maintain registries, operate the sandbox, run inspections and impose administrative measures. The Act creates transitional rules for system registration, conformity assessment timelines and the staged applicability of certain EU AI Act chapters—these align with EU phasing (some obligations fully apply from August 2, 2025/2026 depending on the EU timeline) and Hungarian operational dates (Act: effective 2025‑12‑01 for most provisions; specific sandbox/test‑environment rules to follow). For the parliamentary bill and implementing documents see the official parliamentary PDF: Act LXXV text.

Monitoring and Evaluation

The Act requires national authorities to maintain registries, perform post‑market surveillance, collect incident reports and serious‑incident notifications, and to make periodic evaluations of market behaviour and compliance trends. The Hungarian Artificial Intelligence Council is charged with monitoring the evolution of AI technologies and advising on policy adjustments and guidance. The market surveillance authority must report to the EU single point of contact and cooperate with other Member State authorities; it will also publish annual enforcement and monitoring reports and maintain channels for complaints and whistleblowing.

Penalties, Liability, and Appeals

The Act empowers national market surveillance authorities to apply the full range of EU AI Act enforcement remedies: warnings, corrective orders, product removal or suspension, administrative agreements and administrative fines up to the maxima set out in the EU AI Act (tiered fines for prohibited practices and other breaches). The Act also preserves civil liability routes under national tort and contract law and affords procedural rights and judicial review for parties subject to administrative sanctions. The EU AI Act's fines framework (up to €35 million or 7% of global annual turnover for the most serious breaches; up to €15 million or 3% for other major breaches; lower tiers for supplying misleading information) will inform Hungarian sanctioning decisions and ceilings.

Relationship to Other Instruments

The Act operates alongside the EU AI Act (which remains directly applicable), the national data protection framework (GDPR & the National Data Protection Act; the NAIH remains a key cooperating body), sectoral supervisory regimes (e.g., MNB for finance, sectoral market surveillance authorities), consumer protection and competition law, and national strategies and government decisions establishing the regulatory sandbox. The Act instructs cooperation and division of competences where intersection occurs (for example between NAIH and market surveillance authorities or between sectoral regulators and the AI market surveillance authority).

International Alignment

The Hungarian law explicitly seeks full alignment with the EU AI Act and to implement the Commission's framework in a manner consistent with EU supervisory cooperation. The Act provides for Hungarian participation in EU‑level processes (single point of contact liaison, cooperation with the EU AI Office) and establishes mechanisms for cross‑border enforcement and information sharing. The design also recognizes international accreditation and conformity assessment standards and signals intent to align with European and international standardisation work and the Commission's GPAI (general‑purpose AI) policy workstreams.

Implementation Timeline

Sources and References