A and sophisticated fraud: Lithuania's mission to a safer digital society

Digital technologies are no longer a futuristic luxury: they are part of the skeleton of public services, banking, health and communication. In countries like Lithuania, this reality is particularly palpable: from electronic signature to clinical records, daily life depends on connected and secure systems. But as innovation accelerates, risks also grow; today cybersecurity is both a technical challenge and a matter of social trust and governance.

This diagnosis is the basis for a national initiative funded by the State and coordinated by Innovation Agency Lithuania which is committed to converting university knowledge into high-value commercial solutions to increase the country's digital resilience. Among the most relevant projects is the mission "Safe and Inclusive E-Society" led by the Kaunas University of Technology (KTU) a consortium that integrates universities, cybersecurity companies and research centres with the aim of bringing prototypes to real environments - public institutions and critical infrastructure operators - and thus close the gap between research and practice.

The work being financed covers areas that are already part of the global risk map: from smart buildings that are adapted and learned, to artificial intelligence-based defenses to protect financial services and fintech platforms. Sensors are also developed for the early detection of threats in industrial infrastructure and hybrid incident management systems applicable to public safety, education and enterprise. Other efforts seek to identify disinformation campaigns and automate real-time threat intelligence with smart platforms.

A profound change in this landscape has been the arrival of generative models and large language models. According to Professor Rasa Brūzgienė of the Department of Computer Sciences of KTU, these tools have transformed the logic of fraud: where it was first enough to detect repeated patterns, attackers can now create coherent, well-written and personalized messages that imitate the institutional style. That makes many scams into communications that at first sight seem legitimate.

The risk is not just theoretical. The tools that facilitate more and more sophisticated attacks are available to the public: advanced language models (commercial and open source), voice cloning services and image and video generation solutions that make it possible to create convincing faces and sequences. Research projects such as VALL-E or commercial platforms such as ElevenLabs have shown what voice synthesis can achieve; for its part, ecosystems like Stable Diffusion, DALL · E or publicly available deepfake techniques reduce the barrier to create false images and videos. The technological press has also documented malicious variants focused on facilitating fraud, such as the well-known tools that emerged with commercial names that are abuse-oriented.

The most worrying thing is multimodal orchestration: frames and false faces, documents with altered metadata, texts generated by LLMs, cloned voices and automated agents that complete the work. Together, these components can mock automatic verifications and also deceive people, because the scam no longer "seems" a fraud, but a plausible interaction with apparent backing.

Artificial intelligence also drives a new generation of social engineering. Bots that track social networks, professional directories and filtered databases build detailed profiles of potential victims; then, a LLM adapts messages in real time, changes channels and tone, and responds to doubts using public information or internal procedures precisely cited. Within minutes a single automated campaign can scale and personalize for hundreds of goals, multiplying the effectiveness of deception.

Faced with this scenario, Lithuania shows both challenges and responsiveness. Its digital ecosystem, with a centralized electronic identity and advanced digital government services, gives it both vulnerabilities to protect and advantages for implementing countermeasures at the national level. The country participates and cooperates with European and allied organizations to strengthen its defence, and that public policy has integrated the use of IA in cyberdefence reflects a strategic approach that combines automated detection and resilience efforts.

Combating these threats requires coordinated action on several fronts: developing and deploying detection that goes beyond static patterns, incorporating multimodal behavior and correlation analysis, strengthening human verification processes with tools that detect synthesis devices, and designing protocols that limit damage when an intrusion occurs. In addition, the continuing training of public and private staff and the digital literacy of citizens are equally essential to ensure that confidence in electronic services is not eroded.

The central lesson is that artificial intelligence is not only the lever of the attacker: it is also the best defence tool if used in collaboration between universities, companies and administrations. Initiatives such as the KTU and the Innovation Agency seek precisely to turn scientific potential into services and pilots that are tested in real conditions, thus strengthening citizen confidence in the digital economy.

If the risk analysis and the European response are to be further developed, the reports of specialized bodies should be reviewed: ENISA has documented the evolution of threats related to IA, and the Europol offers evaluations of organized crime on the network. The official channels of the Lithuanian consortium and the universities involved are also useful to follow how research is translated into concrete solutions.

In short, the challenge is clear: the same technology that multiplies productivity and social closeness can be used to undermine public confidence. The answer goes by joint work, investment in applied R & amp; D and continuing education so that the network is safe and accessible to all, and that social and technological innovation always go hand in hand with security.