Lithuania - AI Use in Public Sector

Overview

The Seimas resolution "On the Principles of Use of Artificial Intelligence Technologies in the Public Sector" establishes a non-binding but politically authoritative framework guiding how Lithuanian public-sector bodies should adopt and operate AI technologies. Adopted on 9 May 2024, the resolution emphasises responsible deployment, requiring continual human oversight of administrative decisions, public disclosure when AI affects outcomes, non-discrimination and protection of human rights, and the preservation of service quality regardless of whether tasks are automated. The resolution instructs the Government to implement these principles across ministries, agencies and municipalities and to propose legislative or administrative measures where gaps are identified. The official text can be consulted on the Seimas portal at Seimas: Resolution on the Principles of Use of Artificial Intelligence Technologies in the Public Sector, with contemporary reporting available at LRT article.

Definitions

The resolution uses operational definitions to guide public bodies. "Artificial intelligence technologies" is used broadly to cover algorithmic and machine-learning systems, automated decision-support tools, and other data-driven models that materially influence administrative outcomes. "Public-sector subject" or "public-sector body" refers to state institutions, ministries, government agencies, municipal administrations and other bodies exercising public functions. "Human oversight" denotes meaningful human review and final decision authority for administrative acts materially affected by AI. "Transparency" covers both public disclosure of AI use and internal traceability and documentation sufficient for audit and redress.

Governance and Institutional Framework

The resolution sets out a governance architecture based on shared responsibilities: the Seimas provides political direction; the Government is tasked with coordinating implementation and proposing legislative adjustments; individual ministries and agencies are responsible for operational compliance within their competences; and national supervisory authorities (notably the State Data Protection Inspectorate) retain their statutory powers over data protection and processing. The resolution calls for inter-institutional cooperation and capacity-building measures to ensure public bodies have the skills to assess risks, perform impact assessments and manage procurement or procurement-like evaluations of AI systems. The text asks the Government to consider establishing or designating specialised coordination functions to advise and monitor AI deployment in the public sector; see the official resolution text on the Seimas portal at Seimas resolution (official) and reporting summarising the Seimas vote at LRT.

Key Focus Areas

The resolution highlights a set of substantive focus areas public bodies must prioritise. First, human rights and non-discrimination: AI must not produce outcomes that disadvantage persons on the basis of gender, race, nationality, language, origin, social or economic status, sexual orientation, education, religion or political belief. Second, human oversight and accountability: administrative decisions materially affected by AI require documented human review and clear assignment of responsibility for decisions and consequences. Third, transparency and disclosure: public-sector entities must publish information about AI use, including whether a decision was influenced by AI and the general purpose of the system, while respecting security and legitimate confidentiality. Fourth, data protection and privacy: AI deployments must be compliant with data protection frameworks (GDPR and national implementing law) and the State Data Protection Inspectorate's supervisory powers remain applicable. Fifth, continuity and quality of service: the resolution requires that AI-driven outcomes maintain equal quality of public services compared to human-performed activities. Sixth, risk management and impact assessment: a risk-based approach should be applied, with higher scrutiny for systems that potentially affect fundamental rights or essential services. Finally, cybersecurity and model security: public-sector deployments must adopt reasonable technical and organisational measures to ensure model integrity, prevent data breaches, and guard against manipulation or adversarial attacks.

Implementation Framework

Although politically binding, the resolution leaves detail to administrative action. It instructs the Government to develop implementation guidance, to coordinate capacity-building, and to assess whether targeted legislative changes are necessary to provide enforceable obligations or sanctions. Public-sector organisations are expected to integrate AI governance into existing compliance regimes (procurement rules, data protection processes, administrative procedure, and sector-specific legislation). The resolution encourages the use of documentation, impact assessments (including privacy and fundamental-rights assessments), maintenance of logs for traceability (audit trails of inputs/outputs and decisions), and routine external or independent review where high risks are identified. The State Data Protection Inspectorate and other supervisory bodies are referenced as the competent authorities for relevant enforcement domains; institutions are asked to consult them during design and deployment phases.

Monitoring and Evaluation

The resolution requires that the Government and public-sector institutions monitor AI deployments and report periodically on compliance with the principles. Monitoring should include regular reviews of systems that materially affect individuals or public services, metrics for impact (including measures of fairness, accuracy and service continuity), and public reporting to allow civic oversight. The resolution asks the Government to propose mechanisms for ex post evaluation and to consider public registers of AI systems used in the public sector to improve transparency and enable market and supervisory scrutiny.

Penalties, Liability, and Appeals

The Seimas resolution itself does not create a novel enforcement regime or new sanctions; instead it points to existing legal and administrative accountability channels. Where personal data protection issues arise, supervisory actions and fines available under the GDPR national implementation and the State Data Protection Inspectorate's powers remain applicable. Sectoral laws and administrative procedures can provide remedies, appeals and liability paths for affected persons. The resolution also emphasises institutional responsibility: public bodies remain accountable for consequences arising from AI use and must ensure access to effective redress for citizens.

Relationship to Other Instruments

The resolution is intended to complement and orient application of existing national and EU legal instruments. It references the primacy of human rights and data protection frameworks (EU GDPR and national laws) and invites alignment with forthcoming EU-level AI rules and recommendations. It positions the resolution as a political and administrative bridge while formal legislative measures are considered or enacted to regulate specific technical, market or liability aspects of AI.

International Alignment

In framing principles for national public-sector AI use, the Seimas resolution recognises the need to align with EU and international developments. The resolution calls for coherence with EU instruments (notably the GDPR and the EU-level AI regulatory initiative) and encourages cooperation with EU bodies and regional partners on standards, conformity approaches and resilience against cross-border cyber risks. It also asks Lithuanian authorities to monitor EU-level legal developments and to adjust national practice accordingly.

Implementation Timeline

Sources and References