National Plan for the Development of Artificial Intelligence for the Period until 2032

In healthcare, large quantities of data on patients, medical histories, diagnostics, therapies and the organisation of health services are collected and processed every day. For this reason, healthcare is one of the areas with the greatest potential for the application of AI, but also an area in which errors can have serious consequences. Development must at the same time address legal, ethical, organisational and security issues, in particular issues of privacy, data quality, clinical validation, accountability and the retention of professional oversight by healthcare professionals.

Croatian healthcare institutions already have solutions that use AI models to a certain extent, most commonly as a functional part of certified medical devices or of pilot projects of individual institutions. Investments in digital diagnostic imaging, hospital information systems, telemedicine, robotic surgery, oncology data and the modernisation of operating and diagnostic units are creating the basis for wider and more systematic application.

• 4.4.1.Prevention, early detection of disease and personalised diagnostics+

  The Republic of Croatia will develop and introduce AI systems that support radiology, pathology, cytology, laboratory medicine and other diagnostic fields. Digital radiology systems, modern magnetic resonance imaging and computed tomography devices, and angiographic and endoscopic systems enable the application of models for the analysis of medical images, the detection of suspicious changes and the prioritisation of findings. Pilot projects analysing computed tomography (CT), magnetic resonance (MR) scans and mammography, as well as the use of specialised systems in screening programmes, indicate a possible direction for further development.

  In oncology, standardised and structured findings, the National Oncology Network and national databases will enable the development of models for monitoring health trajectories, assessing response to therapy and personalised treatment support. In robotic surgery and radiotherapy, AI can be used for more precise planning of procedures, contouring and adaptive therapy planning, whereby oversight, decisions and the performance of procedures remain in the hands of licensed healthcare professionals.

  The aim is to gradually increase the availability of clinically validated AI-assisted diagnostics in areas where it can increase the accuracy of findings, shorten processing times and reduce the number of repeated tests. Introduction will be based on evidence of clinical benefit, regulatory compliance and an assessment of the costs and benefits for the health system.

• 4.4.2.Capacity planning and management of healthcare processes+

  Advanced analytical models will be developed for forecasting the consumption of medicines, the need for hospital beds, the workload of emergency services, the availability of medical staff and other key capacities. Digitalised hospital and emergency admissions, monitoring of system performance, the functional integration of hospitals and the development of day hospitals and day surgery are creating the data basis for better planning of the healthcare network.

  Systems for monitoring and preventing medicine shortages and the digitalisation of the medicine pathway and of unit-dose distribution will enable AI analytics of consumption, forecasting of supply disruptions, optimisation of stocks and the detection of inconsistencies in therapeutic processes. Connection with the Central Health Information System (CEZIH) and telemedicine systems can improve the availability of specialist services, reduce waiting lists and enable better planning of primary healthcare resources.

  Tools will also be developed to reduce the administrative burden on healthcare professionals, including Speech-to-Text models for drafting findings, discharge letters, medical histories and other medical documents. Such tools must be adapted to Croatian medical terminology, integrated into secure environments and subject to mandatory verification by a healthcare professional before the document is used.

• 4.4.3.Drug research and analysis of clinical and genomic data+

  The Republic of Croatia has the potential to use AI in the analysis of clinical databases, oncology registers, biomedical images and genomic data, and to develop models that assess therapeutic outcomes, support the identification of therapeutic targets and enable new uses to be found for existing medicines. The automated collection of structured data from hospital information systems will be an important prerequisite for translational medicine, clinical pharmacology and drug research.

  Existing innovation ecosystems, including the AI4Health.Cro EDIH, can support proof of concept, test-before-invest, linking healthcare institutions with researchers and enterprises, and the development of tools for the analysis of clinical pathways, process optimisation and pharmacotherapeutic analytics. Such activities link the area of health with the S3 priority Personalised Healthcare and with the development of digital products and platforms.

• 4.4.4.Secure and interoperable use of health data and AI solutions+

  Strict procedures will be established for quality control, clinical validation and monitoring of medical AI systems, as well as standards of interoperability between health information systems. The processing of health data will be carried out in accordance with the regulations on data protection and medical devices, the AI Act and Regulation (EU) 2025/327. Particular attention will be paid to cybersecurity, access management, monitoring of data use and ensuring that AI is used as a decision-support tool and not as an autonomous substitute for a healthcare professional.

  In the first implementation period, a national secure environment for the processing of health data will be established on the CDU as shared infrastructure for the secondary use of data and for future common European data spaces. It will create the preconditions for the lawful and controlled development, testing and validation of AI solutions. Additional clinical and organisational applications will be introduced gradually, according to the readiness of healthcare institutions, data quality, regulatory requirements and secured funding.

The application of AI in industry is key to increasing productivity, modernising production, using energy and materials more efficiently and developing higher value-added products and services. For the Croatian economy, in which SMEs predominate, a particular challenge is to enable access to the expertise, data, computing capacities, testing environments and funding needed for the transition from individual pilot projects to application in a production environment.

The industrial application of AI builds on investments in digitalisation, automation and robotisation and links several S3 thematic areas. AI can support the transformation of traditional manufacturing activities, but also the development of domestic digital products, embedded systems, specialised software, sensors, semiconductor technologies and industrial platforms intended for the domestic and international markets. In drawing up the National Plan, the draft National Plan for the Development of Industry of the Republic of Croatia for the period from 2027 to 2034 was taken into account in order to align the priorities for the development and application of AI with the future directions of national industrial policy.

• 4.5.1.Smart, flexible and safe production+

  In production processes, AI will be used to analyse data from machines, sensors and production lines, detect deviations early, optimise production parameters and carry out predictive maintenance of equipment. Computer vision will enable automated quality control and defect detection, and digital twins will enable the simulation of production scenarios before they are applied in a real environment.

  AI will support robotics and human–machine collaboration in tasks that are repetitive, physically demanding or hazardous. Application must be focused on increasing the safety and quality of work, with appropriate adaptation of workplaces, the development of employees' competences and clear management of accountability. In industrial and operational technology environments, cybersecurity must be built into the architecture of solutions from the outset, especially when production systems are connected to suppliers, the cloud or remote maintenance services.

• 4.5.2.Efficient use of energy, materials and resources+

  AI models will enable more precise monitoring and optimisation of the consumption of energy, water, raw materials and other resources in production. Analysis of production parameters can reduce scrap, waste and unplanned downtime and improve equipment utilisation. Linking industrial data with energy and environmental indicators will support carbon footprint management, the application of circular economy principles and compliance with the increasingly complex sustainability requirements of customers and value chains.

  Particular potential exists in energy-intensive activities, the processing of food, wood and materials, and production in which small process improvements can bring significant savings. AI can also help in planning production batches, combining orders and scheduling equipment in order to reduce costs, lead times and resource consumption at the same time.

• 4.5.3.Supply chains, industrial logistics and resilience+

  Supply chain management increasingly relies on timely data on demand, stocks, transport, suppliers and external risks. AI will be used for demand forecasting, optimisation of stocks and procurement, detection of disruptions, assessment of the reliability of supply routes and better linking of production, warehousing and transport. This increases the resilience of enterprises to seasonality, geopolitical and market disruptions and sudden changes in the prices and availability of raw materials.

  The development of product and material traceability, especially in the food, pharmaceutical, wood and other regulated industries, will enable faster identification of quality problems and more reliable proof of origin and sustainability. Standardised data and interoperability between enterprises, public registers and common European data spaces will be important for such applications.

• 4.5.4.Development of higher value-added products, services and business models+

  AI will be used in product design and development, simulations, virtual prototyping, personalised and small-batch production and the integration of digital services into physical products. Particular potential lies with Croatian enterprises that develop industrial software, analytical platforms, sensor systems, computer vision solutions, robotics, embedded AI systems and specialised models for individual industrial fields.

  Linking AI with advanced electronics, semiconductor technologies, edge computing and industrial data spaces can open up opportunities for participation in European value chains. Such development should be focused on products and services that can be scaled and exported, and not only on the one-off digitalisation of an individual process.

• 4.5.5.Support for industrial application and for the transition from pilots to scaling+

  Enterprises will be provided with expert and advisory support, test-before-invest, access to research and computing infrastructure, cooperation with research organisations and financial instruments adapted to different stages of development. DIHs and EDIHs will play an important role in assessing digital maturity, selecting use cases, carrying out proofs of concept and linking with technology providers.

  The First Action Plan establishes a horizontal portfolio of instruments for research, development, innovation and technology transfer, targeted vouchers for SMEs, funding for start-ups and scale-ups, loan and guarantee instruments, GovTech and innovative public procurement, and the involvement of Croatian stakeholders in strategic European projects and value chains for AI and semiconductor technologies. These instruments are not limited to a single industry, but will enable the development of projects in activities where there is a clear business need, data and technological readiness and the possibility of a measurable increase in productivity and competitiveness.

AI is becoming an important tool for protecting information systems, critical infrastructure and the population, but at the same time it creates new vulnerabilities and can amplify existing threats. The national approach must therefore simultaneously cover the use of AI for security, the security of AI systems and protection against the malicious use of AI. The development and application of AI solutions in this area will be carried out within the existing European and national cybersecurity and resilience framework, in particular the NIS 2 Directive, the Cybersecurity Act (Croatian), Directive (EU) 2022/2557 and the Cyber Resilience Act, in accordance with their scope. This area is directly linked to the S3 priority Security and Dual Use – Awareness, Prevention, Response and Recovery, as well as to the specific objectives of the National Plan relating to infrastructure, data, legality and competences. Cybersecurity activities related to the development and application of AI will be aligned with the Cybersecurity Act (NN 14/24), the National Cybersecurity Plan from 2027 to 2031 and the associated action plan, in order to ensure the complementarity of measures and avoid their overlap.

• 4.6.1.AI for cybersecurity and the protection of critical infrastructure+

  AI is becoming an important tool for protecting information systems, critical infrastructure and the population, but at the same time it creates new vulnerabilities and can amplify existing threats. The national approach must therefore simultaneously cover the use of AI for security, the security of AI systems and protection against the malicious use of AI. The development and application of AI solutions in this area will be carried out within the existing European and national cybersecurity and resilience framework, in particular the NIS 2 Directive and the Cybersecurity Act (Croatian), Directive (EU) 2022/2557 on the resilience of critical entities and the Critical Infrastructure Act and, in relation to products with digital elements, the Cyber Resilience Act. This area is directly linked to the S3 priority Security and Dual Use – Awareness, Prevention, Response and Recovery, as well as to the specific objectives of the National Plan relating to infrastructure, data, legality and competences.

  Particular attention will be paid to critical infrastructure systems in energy, transport, health, water supply, communications and public digital services. In operational technology systems, including Supervisory Control and Data Acquisition (SCADA) environments, AI can help detect anomalies and provide early warning of failures or security incidents. Such AI solutions must be resilient, segmented and capable of continuing safe operation in the event of a communication outage or the compromise of part of the system.

• 4.6.2.Public security, crisis management and resilience+

  AI can support police, border, customs, fire-fighting, civil protection and other security services in analysing large quantities of data, recognising patterns, managing reports, assessing risks and deploying resources. Possible applications include the analysis of reports and calls, support for operational planning, the detection of incidents in traffic and public spaces, search and rescue and faster assessment of the consequences of fires, floods, earthquakes and other disasters.

  Application in police and other security work will be carried out exclusively on a clear legal basis, with strict respect for the principles of necessity, proportionality, protection of personal data, prohibition of discrimination and human oversight. AI systems will not replace authorised officers in making decisions that produce legal effects, but will serve as support for analysis and operational work. Risks related to biometrics, profiling and the processing of sensitive data will be specifically assessed.

• 4.6.3.Security and resilience of AI systems+

  AI systems may be exposed to manipulation of input data, poisoning of training data, model theft, unauthorised modification, leakage of confidential information and the compromise of the software or hardware supply chain. For this reason, security will be built into the entire life cycle of the system: from the selection and preparation of data and the development and testing of models to deployment, monitoring, maintenance and decommissioning. Particular attention will be paid to protecting the integrity, confidentiality, quality and traceability of data used for training, validating and testing models, and to preventing unauthorised modification of data, models and their key components.

  For systems of public and security interest, rules will be established for identity and access management, activity logging, separation of development and production environments, vulnerability management, security testing and incident response. Business continuity plans, backups, the ability to roll back to a previous version of a model and clearly assigned responsibility for security decisions are also needed. The security and operational risks of AI systems will be continuously monitored and assessed throughout their life cycle, including changes in performance, the emergence of new vulnerabilities and risks arising from dependence on external components and services. In the procurement of models and services, the reliability of suppliers, the location of data processing, dependence on external platforms and the possibility of independent oversight will be assessed.

• 4.6.4.Threats amplified by the use of AI and protection of the information space+

  Generative and other AI tools can be misused for many and currently unforeseeable malicious and illegal activities, including automated cyberattacks, sophisticated (spear) phishing campaigns, fraud, identity theft, the creation of malicious content and the tailoring of attacks to individual users or organisations. Synthetic audiovisual content and deepfake technologies increase the risk of identity manipulation, reputational damage, financial fraud and the undermining of the integrity of the information space, and potentially not only of the security but also of the stability of the Republic of Croatia and its independence and democracy, as well as its constitutional and legal order.

  The response will include strengthening capacities for detecting synthetic content and patterns of coordinated malicious activity, developing protocols for verifying authenticity and rapid response, cooperation between public bodies, research organisations, the media and the private sector, and increasing citizens' AI and media literacy. In doing so, freedom of expression should be preserved and measures that would restrict lawful communication without a clear legal basis should be avoided.

• 4.6.5.Dual use, export control, European and allied cooperation+

  AI technologies, sensors, semiconductors, robotics, autonomous systems and advanced analytics often have civilian and security, i.e. dual, uses. The Republic of Croatia will encourage the responsible development of such technologies, the linking of research and industrial capacities and participation in European programmes and value chains, while complying with the regulations on export control, the protection of classified information and other applicable security requirements. Dual-use items, including, where applicable, software and technology related to AI, are subject to Regulation (EU) 2021/821 of the European Parliament and of the Council of 20 May 2021 setting up a Union regime for the control of exports, brokering, technical assistance, transit and transfer of dual-use items (recast), as well as to obligations arising from the international export control regimes in which the Republic of Croatia participates. Certain AI models, software and associated technology may, depending on their characteristics and purpose, fall within that regime, which is why the research community and enterprises will be provided with timely information on applicable obligations.

  European cooperation in the field of advanced technologies and the defence industry is also pursued through European Union instruments. The Republic of Croatia will support the involvement of domestic research organisations and enterprises in the European Defence Fund, established by Regulation (EU) 2021/697 of the European Parliament and of the Council of 29 April 2021 establishing the European Defence Fund and repealing Regulation (EU) 2018/1092, and in the European Defence Industry Programme, established by Regulation (EU) 2025/2643 of the European Parliament and of the Council of 16 December 2025 establishing the European Defence Industry Programme and a framework of measures to ensure the timely availability and supply of defence products (EDIP Regulation). Participation in these instruments will be developed in line with the European Defence Industrial Strategy (EDIS) and national priorities for the development of defence and technological capabilities.

  As a member of the North Atlantic Treaty Organization (NATO), the Republic of Croatia participates in relevant allied frameworks in the field of security and technological cooperation, including NATO's Artificial Intelligence Strategy, adopted in 2021 and revised in 2024, and the six principles of responsible use of artificial intelligence in defence – lawfulness, responsibility and accountability, explainability and traceability, reliability, governability and bias mitigation. Cooperation is also pursued through the Data and Artificial Intelligence Review Board (DARB), while the Alliance's innovation mechanisms, including DIANA and the NATO Innovation Fund, can support the linking of Croatian research, innovation and economic capacities with the allied innovation ecosystem.

  The military use of artificial intelligence is not the subject of this National Plan. Issues concerning the development and use of AI exclusively for military, defence or national security purposes, to which, in accordance with Article 2(3) of the AI Act, that Act does not apply, are governed by a separate strategic and regulatory framework and fall within the competence of the authorities responsible for defence and national security, in particular in accordance with the National Security Strategy of the Republic of Croatia (NN 73/17), the Defence Strategy of the Republic of Croatia (NN 104/25) and other relevant documents. The Republic of Croatia follows the development of international legal and policy frameworks for the responsible military use of artificial intelligence and autonomous systems, including the work of the Group of Governmental Experts within the framework of the Convention on Certain Conventional Weapons, discussions in the United Nations General Assembly and other relevant international initiatives.

  In the first implementation period, the emphasis is on raising the level of cybersecurity of the national AI infrastructure and implementing selected AI solutions in operational police processes, with the horizontal contribution of measures on data governance, the national framework for AI governance and the regulatory sandbox. Further measures for the protection of critical infrastructure, crisis management, the resilience of AI systems and countering threats amplified by the use of AI will be elaborated through subsequent action plans in cooperation with the competent authorities.

Agriculture, forestry, fisheries, veterinary medicine, hunting, and the food and wood industries have a large amount of production, environmental, spatial and administrative data. AI can contribute to more precise management of natural resources, optimisation of production, reduction of costs and losses, food safety and greater resilience to climate and market risks. In these activities, it is important that AI solutions are practical, accessible and adapted to small producers and different regional conditions.

The area is directly linked to the S3 priorities Sustainable and Circular Food and Customised and Integrated Wood Products, while digital platforms, sensor systems and analytics form its horizontal technological basis.

• 4.7.1.Precision agriculture and management of crops and resources+

  AI models will be developed for the precise management of vegetation, pests, diseases, water resources and the use of fertilisers and plant protection products. Existing systems, including eSavjetnik and AKIS, which bring together agro-climatic data, warnings and expert information, can be upgraded with data on soil, historical yields, producer records, satellite images and sensor measurements. This will enable personalised recommendations, analysis of multi-year trends and timely warnings of risks.

  A smart agriculture platform should link data on land, meteorological conditions, pests, plant protection products and production, and enable their secure use in analytics and AI models. Digital twins of parcels, plantations or production systems can support the simulation of production scenarios, irrigation planning and the assessment of the impact of different decisions before they are applied in the field.

• 4.7.2.Production planning and management of agricultural and biological resources+

  AI will be used to forecast production trends, disease risks, resource needs and the optimal allocation of capacities in crop and livestock production, veterinary medicine, fisheries and aquaculture. Analytics can support the management of state-owned agricultural land, the planning of veterinary activities, animal health monitoring and more efficient targeting of expert and advisory services.

  In fisheries and aquaculture, AI can help in monitoring environmental conditions, estimating biomass, planning feeding, detecting diseases early and managing fishery resources sustainably. In hunting and game management, field data, recordings and spatial data can be combined to monitor populations and reduce damage.

• 4.7.3.Food safety, traceability and sustainable processing+

  AI can improve food quality and safety control, the early identification of risks in production and supply chains, the traceability of raw materials and products, and the optimisation of processing, packaging and distribution. By analysing production and market data, it is possible to plan quantities more precisely, reduce food losses and better link primary production with processing capacities and demand.

  The development of sustainable and circular food systems requires the linking of data on raw materials, by-products, waste and resource consumption. AI can help identify opportunities for reuse and processing, optimise logistics and develop higher value-added products.

• 4.7.4.Forestry, the wood industry and the bioeconomy+

  In forestry, AI will be used for the analysis of satellite and aerial images, forest inventories, monitoring of forest health, early detection of pests and diseases and fire risk assessment. Linking spatial, climate and economic data can improve the planning of felling, regeneration and protection of forest ecosystems.

  In the wood industry, AI can support the optimisation of cutting and material utilisation, automated quality control, the design of customised products, predictive maintenance and the development of new products linked to digital services. This links resource efficiency, design, advanced manufacturing and the S3 priority Customised and Integrated Wood Products.

• 4.7.5.Conditions for wider application+

  Wider application will depend on the quality and interoperability of public and private data, the availability of broadband and sensor infrastructure in rural areas, clear rules on access to data and the trust of producers. Solutions should be easy to use, economically justified and accessible to holdings of different sizes, with the possibility of advisory and technical support.

  The First Action Plan does not contain a separate sectoral measure for agriculture, food and forestry, but enterprises and research organisations from these areas can use the horizontal instruments for research, innovation, vouchers, DIH/EDIH services and funding. Specific sectoral measures will be elaborated once projects have been prepared, lead bodies identified and data and funding secured.

The energy transition, environmental protection and adaptation to climate change rely on large quantities of weather, spatial, sensor, infrastructure and market data. AI can link these sources, enable more precise forecasts, optimise the operation of systems and support earlier identification of risks. At the same time, energy and environmental systems are part of critical infrastructure and require a high level of physical, cyber and operational resilience.

• 4.8.1.Optimisation of generation and integration of renewable energy sources+

  AI models will be introduced that link meteorological data, data from wind farms, solar, hydropower and other systems, and consumption forecasts in order to increase the stability of the electricity system and manage generation from renewable sources more efficiently. Models can simulate generation scenarios, grid oscillations, the operation of microgrids, energy storage and flexible consumption, thereby contributing to better utilisation of renewable sources and security of supply.

  The development of such systems must be accompanied by secure data exchange, network segmentation, access control and monitoring of data integrity. AI models that manage generation or demand flexibility must be resilient to data manipulation, cyberattacks and communication outages, especially in conditions of a high share of decentralised sources.

• 4.8.2.Smart grids, predictive maintenance and energy system management+

  AI solutions will be used for fault detection, anomaly analysis, optimisation of voltage levels, congestion forecasting and intelligent load management. At energy facilities, gas pipelines, oil pipelines, hot-water pipelines and other systems, predictive maintenance models will analyse vibration, temperature, load and other parameters in order to identify faults before more serious consequences arise.

  Predictive analytics can support consumption planning for industrial users, households and the public sector, the management of public lighting and the energy efficiency of buildings, and the development of market and non-market scenarios. The particular value of AI lies in the rapid processing of large volumes of data from SCADA systems, sensors and meteorological sources, which increases the resilience of the system and enables longer-term investment planning.

  The growth of data centres and computing capacities for AI will increase demand for electricity and the need for a stable power supply. The planning of digital infrastructure will therefore be linked with energy efficiency, available grid capacity, renewable sources, waste heat and supply continuity plans.

• 4.8.3.Monitoring of biodiversity, habitats and environmental risks+

  Advanced AI models will be developed for monitoring biodiversity, habitats and environmental pressures, relying on existing databases and systems, including BioAtlas, environmental records, spatial data and remote sensing data. Models can link species observations with the intensity of interventions, research effort, space and time, and present trends through maps, spatial analyses and scenarios of future conditions.

  AI will enable faster identification of unusual phenomena, such as a sudden drop in the number of observations, changes in habitats or growing pressures in a particular area. This can speed up the definition of conservation and restoration measures, improve the assessment of the effectiveness of existing measures and provide a better-quality basis for spatial planning, the management of interventions and investments.

• 4.8.4.Water, waste and the circular economy+

  In water management, AI will be used to optimise water supply networks, reduce losses and carry out predictive maintenance of pumps, valves and pipelines. Linking monitoring, meteorological and sensor data will enable the identification of anomalies, the prediction of failures and the optimisation of energy consumption in near real time.

  In waste management and the development of the circular economy, AI tools can link data from municipal systems, industrial production, monitoring applications and registers. Applications include optimisation of waste collection routes, assessment of material flows, detection of irregularities in disposal, automated sorting and the identification of opportunities for reuse and recycling. Virtual assistants can provide citizens and obliged entities with clear instructions on the correct use of information systems and the fulfilment of obligations.

• 4.8.5.Climate risks, fires and adaptation to climate change+

  Specialised AI models will be developed for forecasting floods, droughts, heatwaves, fires, coastal hazards and other climate phenomena. The models will link data from meteorological and hydrological stations, satellites, environmental sensors, infrastructure systems, agricultural databases and climate scenarios. Automated change detection will enable the identification of early signs of hazards, spatial analysis and the modelling of local and national impacts.

  In fire management, AI can analyse meteorological data, vegetation, satellite images and sensor networks in order to assess risk, detect hotspots and predict the dynamics of spread. Such systems should support operational services, civil protection and local units, with clearly defined protocols for data verification and decision-making.

  No separate sectoral measures for energy, the environment and climate are envisaged in the First Action Plan. Their development in the first period will be supported by horizontal investments in infrastructure, data, cybersecurity, research and innovation. Sectoral measures will be included in subsequent action plans according to the readiness of projects, their strategic importance and the availability of funding.

The transport system covers road, rail, maritime, inland waterway and air transport and generates large quantities of data from cameras, sensors, vehicles, infrastructure, and navigation and control systems. AI can increase safety and traffic flow, improve infrastructure maintenance, optimise public transport and logistics and better connect different forms of mobility. The application is directly linked to the S3 priority Smart and Green Transport.

• 4.9.1.Intelligent traffic management and safety+

  The Republic of Croatia will gradually introduce AI systems for the monitoring, prediction and management of traffic flows on urban and intercity roads. Systems that use cameras and sensors to monitor traffic density, detect incidents and optimise the operation of traffic lights can be upgraded with predictive models that link traffic data, weather conditions, tourism seasonality and real-time events in order to prevent congestion and increase traffic flow.

  AI will support the automatic detection of dangerous situations, such as sudden braking, driving in the wrong direction, moving into a prohibited lane or other risky behaviour, and the sending of warnings to traffic management centres and emergency services. The use of video surveillance and other sensors will be carried out on a clear legal basis, with protection of personal data, system security and limitation of processing to the purpose for which it was introduced.

  In the railway system, AI can be used for monitoring tunnels, bridges, level crossings, power supply systems and critical train components, and for predictive maintenance. In maritime transport and inland navigation, advanced analytics can support navigation monitoring, the identification of dangerous situations, search and rescue and safer management of critical navigation sectors. In air transport, AI can support delay prediction, schedule optimisation and maintenance, in compliance with international safety rules.

• 4.9.2.Public transport and integrated mobility+

  AI will be used to plan more reliable public transport, optimise timetables, adapt routes to actual passenger numbers and dynamically manage the capacity of bus, tram, train and ferry lines. The systems will adapt to seasonality, local travel patterns, island communities and tourist pressures.

  Linking public transport, taxi services, micromobility, rail and ferry lines into integrated mobility systems will enable users to plan journeys more easily and operators to make better use of capacity. In tourist centres, AI can predict traffic peaks, manage parking zones and reduce pressure on city centres.

• 4.9.3.Smart logistics, seaports, inland ports and multimodal transport+

  AI models will be used to optimise routes, reduce empty running, predict delays and manage warehouses using smart sensors and automated processes. Linking data from ports, rail terminals, road systems and logistics centres will enable more precise planning of multimodal freight transport and better connectivity of logistics hubs.

  In seaports and inland ports, AI can be used for the monitoring, control, planning and management of traffic within the port area, the optimisation of the occupancy of berths, storage space and railway tracks, and the prediction of congestion. A precondition is the digitalisation of processes and appropriate equipment with video surveillance, computing and communication systems. In inland waterway transport, AI can support the management of port infrastructure, maintenance, safety and the planning of passenger and nautical traffic.

  In the field of electromobility, AI will be used to forecast energy consumption, plan the locations of charging stations and manage the load on the electricity grid. This links transport planning with the energy system, energy prices and the local needs of users.

• 4.9.4.Predictive maintenance of transport infrastructure+

  AI can automate the assessment of the condition of road pavements, bridges, and railway and port infrastructure on the basis of images, sensor data and maintenance records. Predictive models will enable earlier identification of deterioration, better prioritisation of investments and a reduction in the costs of unplanned interventions. The results should be linked to clear inspection standards and confirmation by expert services.

  The First Action Plan includes a pilot application of AI for the smart management of passenger and municipal ports on inland waterways and a system for the automatic assessment of road pavement condition and predictive maintenance management of county and local roads. These pilot projects should generate knowledge, data models and standards which, after evaluation, can be adapted to other operators and types of infrastructure.

Tourism is one of the strategic sectors of the Croatian economy and an area with great potential for the responsible and targeted application of AI. Application will be focused on increasing the competitiveness and added value of tourism, year-round and regionally balanced tourism, more efficient management of tourist destinations and resources, personalisation of tourism and related services, and strengthening the sustainability and resilience of the sector to external shocks.

Tourism is viewed as a horizontal sector that relies on transport, energy, environmental protection, spatial planning, culture, gastronomy and public data systems. Such an integrated approach enables more efficient management of tourist flows, infrastructure and natural resources and the development of smart, resilient and sustainable destinations.

• 4.10.1.Management of tourist flows and destination pressure+

  AI will be used to forecast arrivals, daily and seasonal peak loads, visitor movements and pressures on transport, municipal infrastructure, cultural heritage and natural resources. By linking data on accommodation, transport, events, weather and spending, it is possible to adjust public transport, traffic management, the operation of municipal services and visitor information in good time.

  In tourist destinations, AI can support the spatial and temporal distribution of visits and parking management, direct flows away from the most crowded locations and monitor the effects on the local community. The aim of managing tourist flows is not only to direct visitor movements in space and time so as to achieve a balance between tourist demand, destination capacity and the quality of the visitor experience, but also to balance the economic, environmental and social impacts of tourism.

• 4.10.2.Personalised, accessible and higher-quality tourism services+

  Digital assistants and recommender systems can provide visitors with information on transport, attractions, accessibility, safety and local services, with multilingual support. Of particular value is the increase in accessibility for persons with disabilities and the adaptation of information to the different needs of users.

  Enterprises can use AI for demand analysis, price and capacity management, optimisation of business processes, development of personalised offers and better customer relationship management. Applications must be transparent, based on lawfully collected data and designed so as not to lead to discrimination or non-transparent profiling of users.

• 4.10.3.Sustainable and resilient destinations+

  AI can support the optimisation of energy and water consumption, waste management, early identification of environmental and climate risks and capacity planning during periods of high pressure. Linking tourism data with transport, environmental and energy systems will enable data-driven decision-making and more efficient management of the destination as a whole.

  The First Action Plan does not contain a separate measure for tourism, but certain transport and public sector pilot projects contribute directly to the management of tourist flows and destinations. Additional measures will be developed in cooperation with the competent authorities and local communities once projects have been prepared, data exchange models defined and sources of funding secured.

In the cultural and creative sectors and the media, generative AI is already having a direct impact on the processes of creation, production, distribution and monetisation of content. Workers in the cultural and creative sectors, institutions and citizens are increasingly using AI tools for textual, musical, photographic, audiovisual and multimedia content. AI can increase productivity, open up new forms of artistic expression and facilitate international access to Croatian cultural production, but at the same time it is changing market relations and the position of authors and rightholders.

It is necessary to ensure a strategic, legally clear and institutionally coordinated approach that enables innovation while protecting copyright and related rights, media integrity and the value of human creative work. The area is linked to the S3 priority Digital Products and Platforms.

• 4.11.1.AI in the creation, production, distribution and accessibility of content+

  AI will be used as a tool to support the creation, processing, translation, subtitling, restoration, cataloguing, searching and distribution of cultural and media content. Cultural institutions can use AI for processing digitised heritage, automatic metadata enrichment, content recognition and the development of accessible digital services. Enterprises in the cultural and creative sectors can develop new products, interactive formats and internationally competitive services.

  It is necessary to distinguish clearly between situations in which AI is an auxiliary tool of a human author, with a relevant human creative contribution, and content created without such a contribution. Guidelines for institutions, authors and enterprises should help with the lawful and responsible use of tools, the documentation of the creative process and the provision of information to users.

• 4.11.2.Use of protected content for training AI models+

  Training generative AI models involves processing large quantities of textual, musical, visual and audiovisual content, much of which is protected by copyright. The key issues concern the lawfulness of the use of content, the application of the text and data mining exceptions under Articles 3 and 4 of Directive (EU) 2019/790, the transparency of data sources, the exercise of the right to opt out where provided for by regulations, and licensing models and remuneration for rightholders.

  In the development of national language technologies and the large language model for the Croatian language, particular attention will be paid to the lawful and transparent use of national linguistic and cultural corpora. Digital and linguistic sovereignty must not be achieved to the detriment of domestic authors, publishers and other rightholders. The development of clear guidelines, cooperation with collective management organisations, appropriate licensing models and, where applicable, fair remuneration, as well as technical solutions for the identification and monitoring of the use of protected content, will be encouraged.

• 4.11.3.Generated content, authorship and transparency+

  At the content generation stage, issues arise concerning the imitation of an author's recognisable style, possible unauthorised adaptations, the authorship and legal protection of generated content, misleading attribution and infringement of authors' moral rights. It is necessary to develop regulatory guidelines and institutional capacities for the interpretation and application of copyright law in order to ensure legal certainty for authors, cultural institutions, entrepreneurs and citizens.

  Transparency of content generated using AI or manipulated using AI will be important for the protection of consumers, rightholders and the integrity of the information space. The labelling of content will be carried out in accordance with European regulations and with due regard to the context, purpose and possible risk of deception, and in particular in accordance with the transparency obligations under Article 50 of the AI Act. Particular attention will be paid to deepfake content, identity manipulation and the use of synthetic content in the media and political space.

• 4.11.4.Artistic, market and social dimension+

  AI tools are changing the way in which creativity, authorship and the creative process are perceived and may affect the perception of the value of human work. There is a risk of homogenisation of aesthetic patterns, excessive standardisation of content and concentration of market power on platforms that control AI models and distribution. At the same time, responsible application can open up space for new artistic practices, interdisciplinary cooperation, the preservation of and access to heritage and the internationalisation of Croatian cultural production.

  National policies should at the same time protect the fundamental principles of copyright and the moral rights of authors, ensure transparency where necessary to protect users and rightholders, encourage the development of domestic AI solutions in the creative sector and strengthen the capacities of cultural and media institutions and enterprises for the lawful and responsible use of AI tools.

  The First Action Plan includes a measure for monitoring the development and application of AI in culture and the media and for the timely adaptation of the legislative, institutional and implementation framework to EU standards. On the basis of the results of monitoring the development and application of AI, dialogue with authors, rightholders and cultural and media institutions, and the development of the European framework, additional measures will be elaborated in subsequent action plans.