Romania - National AI Strategy

Overview

The Cadru Strategic Național în domeniul Inteligenței Artificiale (CSN-IA) is a programmatic national framework document (draft version dated 28 July 2023, posted for public consultation in October 2023) developed under the project "Cadru strategic pentru adoptarea și utilizarea de tehnologii inovative în administrația publică 2021–2027" (SIPOCA 704). The document was prepared by the Authority for Digitalisation of Romania (ADR) in partnership with Universitatea Tehnică din Cluj-Napoca and a multidisciplinary group of experts; it provides a strategic roadmap for 2023–2027 to foster AI uptake across public administration, RDI and industry while protecting rights and building trust. The CSN-IA positions Romania to align with EU AI priorities and to prepare operational instruments and legal alignments. See the primary draft: Cadrul strategic național în domeniul inteligenței artificiale — draft (Oct 2023).

Definitions

The draft defines AI broadly as systems that exhibit intelligent behaviour and can take actions with some degree of autonomy. It distinguishes between categories of AI uses (general-purpose models, domain-specific applications, human-in-the-loop systems) and aligns risk-aware terminology with EU practice: low-risk, limited-risk, high-risk (sectoral) and uses requiring specific safeguards. It also defines key governance terms such as "data stewardship", "infrastructure", "conformity" and "monitoring indicators" to provide a common language for stakeholders.

Governance and Institutional Framework

The CSN-IA proposes a multi-level governance architecture centered on ADR as the national coordinator for AI policy implementation, working together with relevant ministries (research, health, education, interior, justice), academic partners (e.g. Universitatea Tehnică din Cluj-Napoca) and sectoral stakeholders. It recommends establishing an inter-institutional coordination mechanism, a national scientific and ethics council for AI, and designated AI focal points within ministries. The framework foresees clear roles for monitoring, standardization, public procurement guidance, and the development of a national registry/inventory of AI systems used in public administration to support transparency and market surveillance.

Key Focus Areas

The CSN-IA concentrates activity into six programmatic objectives: (1) Education and skills—raising baseline digital literacy and creating specialized AI curricula and lifelong learning; (2) Infrastructure and datasets—investment in cloud, high-performance computing, curated datasets and interoperable data governance; (3) RDI and innovation—support for centres of excellence, startups and technology transfer; (4) Partnerships and transfer—public-private-academic collaboration and public procurement of trustworthy AI; (5) Adoption—pilots and scaling of AI in public services and sectoral adoption (health, transport, education, justice, agriculture); and (6) Governance and regulation—preparing regulatory options, conformity mechanisms and alignment with EU rules. Cross-cutting priorities include ethics, data protection, cybersecurity, standards, SME support and inclusion measures.

Implementation Framework

Implementation is envisaged through programmatic measures, specific projects and funding instruments (EU, national and PNRR-linked opportunities). The CSN-IA recommends phased investments, capacity-building calls, establishment of national competence centres and a dedicated implementation secretariat under ADR. It identifies indicators and milestones and calls for sectoral strategies and roadmaps (e.g., for healthcare, education and transport) to operationalize objectives. The draft also proposes templates and operational guidance for public procurement, risk assessment and transparency documentation for deployed AI systems.

Monitoring and Evaluation

The framework prescribes a monitoring and evaluation system composed of quantitative indicators (adoption rates, number of specialists trained, infrastructure capacity, number of compliant projects) and qualitative reviews. ADR, in coordination with partner ministries and the proposed national AI council, would publish periodic progress reports and use a public dashboard for selected indicators. Evaluation cycles and procedures for mid-term corrections are included to ensure adaptability and evidence-based policy adjustments.

Penalties, Liability, and Appeals

CSN-IA is a programmatic draft and does not itself create new administrative sanctions. It proposes that liability, enforcement and penalties for AI systems be dealt with in subsequent sectoral legislation and regulatory instruments, aligned with EU instruments (including the forthcoming EU AI Act) and existing regimes (e.g., GDPR for data protection, consumer protection law, sectoral health or transport rules). The document recommends a pathway for designing conformity assessment, market surveillance and appeal channels at the national level.

Relationship to Other Instruments

The draft maps existing national strategies and European frameworks, and explicitly recommends alignment with the EU Coordinated Plan on AI, the Digital Decade objectives and ongoing EU legislative initiatives. It cross-references national RDI strategies, PNRR commitments and sectoral digitalisation plans to avoid duplication and to ensure complementarity. The CSN-IA identifies legal gaps where follow-on regulation or adaptations of existing laws (data protection, procurement, civil liability) will be necessary.

International Alignment

International alignment is a central aim: the draft instructs Romania to align technical and governance measures with EU standards and to increase participation in EU research infrastructures (EOSC, EuroHPC), to contribute to international standardization bodies and to develop bilateral and multilateral collaboration in AI research and skills development. It also stresses EU conformity approaches and transposition of obligations that may result from the EU AI Act.

Implementation Timeline

Sources and References