Lithuania - AI Regulatory Sandbox Policy

Overview

The Lithuanian AI Regulatory Sandbox Policy creates a supervised national environment for testing innovative AI systems before they are placed on the market. The initiative implements the requirements of the EU Artificial Intelligence Act (Article 57) and positions the national Ministry of the Economy and Innovation and the Innovation Agency (Inovacijų agentūra) as central actors. The sandbox aims to foster innovation, facilitate compliance pathways (particularly for SMEs and start-ups), and support conformity assessment processes while ensuring that risks to health, safety and fundamental rights are identified and mitigated. National authorities will coordinate with the EU AI Office and the AI Act Board, and the scheme is designed to enable cross‑border cooperation where appropriate.

Definitions

Key terms used for the sandbox policy follow the EU AI Act lexicon and national implementing documents: "AI regulatory sandbox" means a controlled environment supervised by a competent authority for the development, training, testing and validation of AI systems under an agreed sandbox plan; "provider/prospective provider" means the natural or legal person offering or intending to offer an AI system; "notified body" refers to entities authorized to carry out conformity assessment of high‑risk AI systems; "market surveillance" denotes the activities performed by the designated national authority (RRT in Lithuania) to verify compliance and carry out corrective actions when necessary.

Governance and Institutional Framework

Lithuania designates the Innovation Agency as the primary competent authority to create and operate the AI sandbox and to act as the national notifying authority for conformity assessment matters; the Communications Regulatory Authority (RRT) is appointed as market surveillance authority and single contact point for market supervision. The Ministry of the Economy and Innovation has led legislative amendments (to the Law on Technology and Innovation and the Law on Information Society Services) to allocate roles, and will coordinate inter-agency cooperation with the State Data Protection Inspectorate and other sectoral regulators. National authorities are required to allocate sufficient resources and to liaise with the EU AI Office/Board as set out in the AI Act.

Key Focus Areas

The sandbox policy focuses on four principal areas: (1) Information and outreach — centralised, one‑stop information on the AI Act and sandbox services made available online and through events; (2) Regulatory guidance and conformity pathways — hands‑on regulatory advice, pre‑certification support and help to prepare documentation for conformity assessment and notified bodies; (3) Supervised testing of AI systems (including staged real‑world tests for certain high‑risk systems with market surveillance oversight); (4) Technical and infrastructural support — access to HPC resources, technical testbeds and expert assessment services. The programme emphasises SME prioritisation (free or low‑cost access), simplified application processes, and the publication of non‑sensitive exit reports to support lawful market access.

Implementation Framework

Implementation is led by the Innovation Agency which will publish operational rules (eligibility, application, selection, confidentiality, data handling, duration and exit procedures). Projects must agree a sandbox plan with the competent authority that sets scope, duration, testing conditions, risk mitigation measures and reporting obligations. The Communications Regulatory Authority will supervise market surveillance aspects and may permit selected testing of AI systems that would otherwise require full conformity assessment prior to market placement, thereby enabling pre‑certification trials under strict safeguards. The national approach integrates data protection supervision for projects handling personal data (consistent with Article 59 of the AI Act) and includes mechanisms for involvement of other competent authorities when sectoral law is engaged.

Monitoring and Evaluation

National competent authorities will maintain oversight through mandatory documentation, incident reporting and tailored monitoring of each sandbox project. Lithuania will submit annual reports to the EU AI Office/Board once the sandbox is established, containing lessons learned, incidents, mitigation outcomes and recommendations for policy adaptation. The Innovation Agency will publish abstracts of annual reports and (with confidentiality safeguards) exit summaries from sandbox projects to foster regulatory learning and share best practices nationally and across the EU.

Penalties, Liability, and Appeals

Providers participating in the sandbox remain liable under national and Union liability law for damages caused during testing. However, the AI Act specifies that if prospective providers adhere to the agreed sandbox plan and follow the competent authority’s guidance in good faith, administrative fines under the AI Regulation will not be imposed for infringements discovered in the sandbox context. National competent authorities retain full supervisory and corrective powers, including the ability to suspend or terminate a testing project where significant risks to health, safety or fundamental rights are identified. The sandbox operating rules will set procedures for appeals and dispute resolution regarding supervisory decisions.

Relationship to Other Instruments

The sandbox policy is explicitly linked to the EU AI Act (Regulation (EU) 2024/1689) and national amendments to the Law on Technology and Innovation and the Law on Information Society Services. It also complements national innovation support programmes and HPC initiatives, and coordinates with data protection law (GDPR) and sectoral regulation (e.g., health, transport) where relevant. The Innovation Agency will also assess and support the accreditation of notified bodies for conformity assessment, aligning national practice with EU delegated and implementing acts.

International Alignment

The Lithuanian sandbox is designed to be interoperable with EU systems and to facilitate cross‑border cooperation among national competent authorities, per the AI Act’s requirement to inform the AI Office and the Board and to enable cooperation. The policy aligns with EU guidance and the EU AI Office’s single interface and reporting structures, and aims to share lessons and test outcomes to contribute to EU‑level regulatory learning and harmonisation.

Implementation Timeline

Sources and References