Notes
4 The situation in Luxembourg
Young people in Luxembourg are navigating a rapidly evolving digital ecosystem shaped by artificial intelligence and pervasive online media. While international studies highlight the impact of screen time and social media on adolescent well-being, national data reveal a sharp rise in AI use among students and young adults, alongside persistent challenges such as cyberbullying, disinformation, and algorithmic manipulation (see Schumacher et al., 2025). These developments underscore a dual reality: AI offers new opportunities for learning and inclusion, yet amplifies existing social inequalities and introduces complex risks to mental health, privacy, and democratic participation. Understanding these dynamics is essential for designing effective prevention strategies and educational frameworks that foster resilience and critical AI literacy.
4.1 Online practices and media consumption
While artificial intelligence is increasingly shaping the digital spaces in which young people interact, learn, and communicate, to our knowledge large-scale studies have yet to investigate its impact. However, the broader effects of intensive online communication and screen exposure on young people’s well-being are already visible in international monitoring data. For instance, the Health Behaviour in School-aged Children (HBSC) studies provide valuable evidence over time on these dynamics across Europe, including Luxembourg.
Digital communication and screen-based activities have become a central part of adolescents’ daily lives in recent years. In 2022, the international HBSC study (Boniel-Nissim et al., 2024) reported that 36% of adolescents across Europe were in continuous online contact with their friends or other people. Girls were more likely than boys to maintain such contact and this gender gap increased with age. The same study shows that gaming is another dominant screen-based activity among adolescents (Boniel-Nissim et al., 2024). Approximately one in five adolescents met the criteria for problematic gaming behaviour, with higher prevalence among boys (26%) than girls (8%), and the proportion of problematic gamers increased with age.
Problematic social media use (PSMU) affected 11% of adolescents across the WHO European Region (Boniel-Nissim et al., 2024). However, contrary to problematic gaming behaviour, it was more prevalent among girls (14%) than boys (8%). The gender gap also widened with age, particularly among 15-year-olds. In Luxembourg, the prevalence of problematic social media usage increased from 5.9% in 2018 to 9.1% in 2022 (Catunda et al., 2024), reflecting a similar growing engagement on online interaction, especially for girls (7% to 12%) (HBSC Luxembourg, 2023). During the same period of time, cyberbullying victimisation also increased from 11% to 13%, with a particularly sharp increase among 13- to 14-year-olds (from 11% to 16%) (HBSC Luxembourg, 2023).
The HBSC 2022 Luxembourg survey also reveals that almost one in five adolescents felt lonely “most of the time or always” in the past 12 months, with a higher prevalence among girls (24.1%) compared to boys (12.0%) and increasing with age (10.3% of 11- to 12-year-olds; 23.7% of 17- to 18-year-olds) (Catunda et al., 2023). These findings align with broader European HBSC results. Across Europe, 15-year-old girls have the highest rates of loneliness: 28% reported feeling lonely, whereas for boys of the same age this proportion is 13% (Cosma et al., 2023). While loneliness was only measured in 2022, a sharp increase can be observed in other mental health indicators. For instance, multiple psychosomatic health complaints (an indicator of mental distress) rose from 36% in 2018 and 44% in 2022 (Cosma et al., 2023); while in Luxembourg, it went from 40.1% in 2018 to 48.8% in 2022 (HBSC Luxembourg, 2023) (Catunda et al., 2023). For girls, the increase was steeper: from about 49.1 % in 2018 to 62.3% in 2022, while for boys it increased from about 31.0 % to 35.4 % in the same period.
The recently published Medialux 2024 report highlights that digitalisation has led to an ecosystem dominated by internet and social media, where young users (18–24) spend up to 5.5 days per week on social platforms and increasingly rely on AI tools (50% use AI weekly) for information (Kies & Lukasik, 2024). This shift raises concerns about cognitive shortcuts: generative AI reduces the effort of cross-checking sources, potentially weakening critical thinking and fostering opinion homogenisation. Social media algorithms amplify personalisation, creating filter bubbles that limit exposure to diverse viewpoints, especially among youth, who are most affected. Combined with the prevalence of disinformation (60% encounter fake news often) and lower verification habits among younger users (34% rarely check facts), these dynamics heighten vulnerability to manipulation. Moreover, influencers play a growing role in shaping opinions (85% exposure among those 18–24 years old), blurring lines between credible and non-credible sources. According to the internal staff survey, advisors report that young people’s engagement with AI systems is often driven by curiosity and shaped by situational impressions such as finding the tools “exciting”, “helpful”, or “personal”. These practitioner observations underscore the need to strengthen critical evaluation skills in AI-mediated information environments.
Empirical findings further show that young people use AI tools in a variety of ways: as an alternative to human support, as linguistic support in Luxembourg’s multilingual context, for entertainment and creative leisure activities, with a focus on time-saving, and for the development of competencies (Langehegermann et al., 2025) — this latter aspect being closely linked to processes of self-directed learning. Accordingly, young people increasingly rely on digital tools for autonomous learning, with younger users (16–20) more frequently turning to AI-based tools and language models, while young adults (21–29) preferentially engage in more structured, certified online courses (Bulut et al., 2025, p. 106). At the same time, the Medialux 2024 report highlights ambivalent effects: while digital environments enable participation, creativity and social connection, they also intensify challenges related to self-regulation, time management, peer pressure and exposure to online risks such as cyberbullying and unwanted sexual contact.
4.2 Digital competence and educational inequalities
Non-formal educational settings are likewise increasingly understood in international research as key contexts for the development of critical (AI-related) media competences, as they are closely tied to young people’s everyday lives, place relationships and trust at the centre, and organise learning processes in more dialogical and participatory ways than many formal settings (Iske & Kutscher, 2020). Studies on (digitalised) youth lifeworlds and on the role of non-formal and informal education indicate that open youth work and youth organisations, in particular, provide important spaces in which young people can reflect on their media practices, ask questions and address inequalities in the access, use and interpretations of digital technologies (Pawluczuk, 2023; Fujii, Hüttmann & Kutscher, 2020; Kutscher et al., 2020).
The qualitative sub-study of the Luxembourg Youth Report 2025 (Biewers et al., 2025 forthcoming) illustrates how non-formal education contributes to preparing young people for dealing with AI: awareness-raising conversations, low-threshold information, thematic projects and pedagogically guided everyday experiences with AI are described by young people as key learning opportunities. Informal “doorstep conversations” about concrete problems or unsettling user experiences strengthen their capacity to reflect on risks, opportunities and courses of action when engaging with AI, while exchanges with peers in these settings are important for negotiating AI-related knowledge, articulating youth-cultural practices and educational concerns around the technology, and learning from one another (Biewers et al., 2025, forthcoming). These findings resonate with recent European work that conceptualises non-formal education and youth work as suitable contexts for building digital and AI-related competences, fostering critical-reflexive attitudes and specifically qualifying pedagogical professionals for AI-related topics (Vermeire et al., 2023; Kechagias, 2025; Council of Europe-Youth Sector, 2023; Biewers, Latz & Weis, 2025). Moreover, the findings of the Luxembourg Youth Report suggest that non-formal educational spaces are also used as places of relaxation and relief from digitality, in which analogue experiences and peer interaction are deliberately preferred (Biewers et al., 2025).
These findings on young people’s mental health and well-being underscore the broader social and educational implications of digitalisation. Increasing screen exposure, online interaction and the emotional pressures of digital life do not only affect adolescents’ health but also intersect with learning opportunities, motivation, and school engagement. Against this background, it is essential to consider how structural inequalities in access to resources, digital competences, and language skills further shape young people’s ability to benefit from digital learning environments. This link becomes particularly evident when examining national education data for Luxembourg.
The National Education Reports for Luxembourg (2015, 2018, 2021, 2024), together with the international studies ICILS (Fraillon, 2024) and PISA (2018), highlight persistent educational inequalities in Luxembourg — both at the primary and secondary level as well as in higher education. Children and young people from socio-economically disadvantaged families, as well as learners who speak a language other than Luxembourgish or German at home, continue to show lower average levels of academic achievement and competence. The most recent ICILS 2023 results further confirm this pattern in the domain of digital competences. Luxembourg ranks among the countries with the largest socio-economic disparities in Computer and Information Literacy (CIL): students from households with a low socio-economic status (ISEI < 50) achieved an average of 472 points, while those from higher-status households (ISEI ≥ 50) reached 531 points — a gap of 59 points, far above the international ICILS average.
Importantly, the ability to identify false or misleading information is an integral part of the competences measured under CIL. In particular, the first domain — “Accessing and evaluating information” — plays a key role in this regard. The ICILS framework indicates that insufficient CIL competences make young people more vulnerable to AI-generated disinformation, algorithmic manipulation and populist narratives, thereby posing a potential threat to informed democratic participation. Strengthening CIL therefore represents a fundamental prerequisite for developing AI literacy, understood as the ability to critically analyse, interpret and use AI-based information and communication systems in a democratic and responsible way. This finding illustrates how strongly digital competences in Luxembourg remain tied to social background, even in one of Europe’s wealthiest and most digitally advanced societies.
This close link between digital competence, critical thinking and democratic participation also determines how effectively young people can engage with artificial intelligence in learning contexts. In particular, those with limited CIL skills are less equipped to use AI tools critically and are therefore more likely to rely on them unreflectively. In this context, AI-supported chatbots can serve as valuable individual learning partners or tutoring systems — provided that students possess the necessary competences to use these tools effectively and critically. This potential is particularly relevant for children whose parents or guardians are unable to support them with schoolwork, either because they do not speak the language of instruction or lack the necessary educational background. For these learners, AI tools can act as accessible learning companions, offering explanations, practice exercises and personalised feedback to help bridge learning gaps at home. As many young people already use AI tools in their everyday lives, the integration of such technologies into formal education makes the development of robust AI literacy indispensable.
The recently published KI-Kompass (SCRIPT, 2025) reflects this understanding and, in line with international research, defines three key learning dimensions: learning without AI, learning with AI, and learning about AI — the latter serving as the foundation for all competent AI use. The findings of the SCRIPT survey on artificial intelligence in education (Summer 2025) complement the quantitative data collected in the Épreuves Standardisées (EpStan) 2023 and 2024, providing current insights into students’ perceptions, usage patterns and attitudes.
While the EpStan results primarily reveal that the proportion of grade 9 students using ChatGPT or other AI tools on their school iPads increased from 18.7% in 2023 to 33.5% in 2024, the SCRIPT data demonstrate that artificial intelligence has now become an established element of everyday school life. According to the survey, around 40% of students use AI tools at least once a week, mostly for supportive purposes such as explaining terms, searching for information, translating, or simplifying texts.
At the same time, the results confirm a continuing ambivalence in dealing with AI: teachers tend to perceive it both as an opportunity and a risk, whereas students are generally more optimistic and assess themselves as significantly more competent in using the technology. Both groups agree that AI-related competencies should systematically become part of the national educational mission—an objective supported by 77% of students and 95% of teachers. The growing usage rates reported in the EpStan surveys can thus be situated within a broader educational policy framework that marks a shift from individual experimentation to an increasingly institutionalised engagement with artificial intelligence within Luxembourg’s school system.
Since the launch of the KI-Kompass, all teachers have gained IAM-based access to the Fobizz platform, which provides data-protection-compliant AI tools, training courses and teaching materials to support the safe, ethical and pedagogically sound integration of AI in teaching practice. By the end of 2025, students are also expected to receive regulated access to these tools, further advancing Luxembourg’s capacity to implement AI literacy at scale.
At the same time, the risks associated with digitalisation and AI continue to evolve. Data from the Youth Survey Luxembourg 2024, the OEJQS Factsheet 01/25 (2025) and the BEE SECURE Radar 2025 indicate that cyberbullying is a widespread and persistent issue among young people in Luxembourg. Results published in the National Youth Report 2025 and based on the 2024 wave of the Youth Survey Luxembourg — a representative online survey with 4,779 fully completed and analysable questionnaires — show that, among primary and secondary school pupils, girls are more frequently victims of cyberbullying, whereas boys are more often reported as perpetrators. Moreover, younger pupils aged 12 to 15 report substantially higher levels of cyberbullying experiences than older adolescents and young adults, experiencing cyberbullying almost twice as often as 16- to 25-year-olds (Bulut et al., 2025, p. 121). The OEJQS Factsheet, based on a representative sample of 1,523 adolescents and young adults aged 12 to 29, found that around one in three respondents (32%) had experienced cyberbullying at least once, with 16% reporting incidents within the past year. The BEE SECURE Radar, drawing on an online survey of 916 young people aged 12 to 30, reports similarly high levels of victimisation: among 12- to 16-year-olds, 44% state that they have experienced cyberbullying at least once, while 30% reported incidents between June 2023 and June 2024 (equivalent to 10% of all respondents in that age group). Among 17- to 30-year-olds, 44% report lifetime experiences of cyberbullying, and 4% report cyberstalking during the same period.
In the emerging era of generative artificial intelligence, these findings acquire new urgency. AI-driven technologies — such as image and video manipulation tools, deepfake generators and synthetic voice cloning — are amplifying the forms and reach of online harassment. Fabricated intimate images, AI-assisted impersonation and algorithmically boosted hate campaigns introduce new dimensions of harm that are harder to detect, trace, and prevent. These developments underscore the growing need to integrate AI-related risks into prevention and awareness programmes.
Against this backdrop, the continuous adaptation of BEE SECURE’s educational and preventive activities is essential. Embedding AI-related risks into its training programmes for children and youth (with a particular focus on 12- to 15-year-olds), as well as for parents and educators, is key to ensuring that all stakeholders can recognise, assess and respond appropriately to AI-mediated threats such as deepfake pornography, synthetic sextortion, or AI-enhanced bullying. By systematically fostering AI literacy, ethical awareness and digital resilience, BEE SECURE can play a pivotal role in preparing young people to navigate an increasingly algorithmic and AI-driven online world safely and responsibly.
The evidence highlights a dual challenge for Luxembourg: while AI-driven technologies offer new opportunities for learning, they simultaneously reinforce existing social and educational inequalities and contribute to rising mental health concerns among young people. Strengthening AI literacy across all age groups, ensuring equitable access to trustworthy technologies, and promoting well-being in online environments are therefore essential steps towards empowering the next generation to participate safely, critically and confidently in a rapidly evolving digital society.
4.3 Vulnerable groups and AI-related risks
In addition to the inequalities within the education system, other imbalances exist in the distribution of risks associated with AI. Vulnerable groups are particularly affected in this regard. Although data on this topic is limited in Luxembourg, some guidelines can be derived from general research.
The literature on digital risks among youth notes that young people, especially adolescents, constitute an especially vulnerable audience because of their high exposure to digital environments and their still-developing capacity to perceive and manage online risks. Similarly, in the AI context, vulnerable populations include groups such as women, racial minorities, refugees, and those with lower socioeconomic status, who are often underrepresented in datasets and are excluded from decision-making processes that shape AI technologies (Shanklin et al., 2022, Green et al., 2024). These vulnerabilities are particularly relevant to young people, who are the most frequent and intensive users of digital technologies. High exposure, coupled with a lower perception of risk and higher trust in online environments, increases their digital vulnerability (Byrne et al., 2016). However, Byrne et al. (2016) also argue that young people’s risk perception is influenced by their online experiences, and they most frequently engage in activities they do not view as risky, a pattern that aligns with Livingstone’s (2011) view that education can help mitigate vulnerability to online risks. In the context of the National Youth Report — and in particular within the qualitative interviews with young people — the material indicates at several points that it is especially the younger adolescents, aged approximately 12 to 14, who engage in an exploratory and experimental use of new media and AI, while paying comparatively less attention to potential risks and dangers. As a result, they can be considered more exposed and vulnerable than older adolescents.
Gender differences and inequalities deepen this vulnerability: while boys consume and are exposed to more sexual content, girls are in general more at risk of being the victims of said sexual content, in situations such as sexting or sexual extortion (Villanueva Blasco & Serrano Bernal, 2019) or pornographic deepfake content (Securityhero 2023). Empirical findings from Luxembourg confirm pronounced gendered inequalities in digital risk exposure: girls are more frequently affected by negative online experiences and unwanted sexual messages. According to the Youth Survey Luxembourg, around 67% of female adolescents report negative online experiences, compared to 61% of male adolescents, and unwanted sexual approaches are reported more than twice as often by girls than by boys (Bulut et al., 2025, p. 117). Drawing on 36 in-depth interviews and digital diaries with young people aged 12 to 29, the National Youth Report 2025 further shows that girls frequently report experiences of exclusion and harassment (Käckmeister et al., 2025), for instance within gaming communities (Langehegermann et al., 2025). The literature confirms that girls also experience more online harassment, social pressure, and anxiety compared to boys (Carvalho et al., 2018). These emotional harms mirror technology-mediated trauma, with some arguing that AI-driven systems, especially generative AI, can inadvertently re-traumatise users by reinforcing bias or misinformation (Abdulai, 2024). Thus, while AI offers opportunities for autonomy and engagement, it also introduces complex psychological and social risks, particularly for digitally active youth.
Socio-economic disparities further shape digital vulnerabilities: adolescents from lower socio-economic backgrounds report negative online experiences more frequently (around 65%) than those from higher-status households (57%) and are also more often confronted with unwanted sexual requests (Bulut et al., 2025, p. 186). Qualitative findings from the National Youth Report 2025 further underline that these risks are unevenly distributed not only in terms of exposure, but also with regard to digital competencies (of both young people and their parents), available coping resources and support structures, especially within families (Meyers et al., 2025, pp. 212). In this context, families who struggle to keep pace with technological developments face particular challenges in relation to AI technologies, limiting their capacity to effectively support and accompany their children in developing meaningful and reflective AI use.
GenAI technologies encode biases from underrepresented datasets, leading to what Holmes (2018) describes as the reinforcement of exclusion and “lack of belongingness” among marginalised users. LLMs can, for example, misgender LGBTQ+ people, stereotype them or limit their expressions (McAra-Hunter, 2024). The reliance on data from WEIRD (Western, Educated, Industrialised, Rich, Democratic) societies (Henrich et al., 2010) means that minority and non-Western populations, including youth from different cultural backgrounds, are often misrepresented or ignored by AI systems (Mihalcea et al., 2025). For young users, who are still forming their identities, such misrepresentation can produce feelings of anxiety, self-inadequacy, and alienation.
Despite these concerns, the literature suggests several pathways to address these vulnerabilities. In the case of youth, enhancing digital literacy and risk awareness can be a promising manner of empowering young users to recognise and respond to online risks more effectively (Ramos-Soler et al., 2018), and some scholars have encouraged participatory design as a manner of combating a lack of representation or misrepresentation (Solyst et al., 2023).
4.3.1 Strategies for supporting vulnerable groups
Building on the findings presented in this chapter, the data from national and international studies reveal that vulnerability in the context of digitalisation and artificial intelligence is multi-dimensional. Socio-economic, linguistic, gender-related, developmental, and psychological factors intersect to shape children’s and adolescents’ ability to participate safely and equitably in an AI-driven society. Certain groups in Luxembourg are particularly at risk of educational exclusion, online victimisation, or digital manipulation. The following table summarises the main vulnerable groups of children and young people in Luxembourg in relation to AI-related and digital risks. It outlines the primary risk factors, the ways in which artificial intelligence may amplify these vulnerabilities, and specific focus areas through which BEE SECURE can address them within its existing and future training, counselling, and awareness programmes.
| Vulnerable Group | Primary Risk Factors | AI-Related Amplification | Recommended Actions |
|---|---|---|---|
| Learners from socio-economically disadvantaged backgrounds | Limited access to devices and guided digital learning; lower CIL competences (ICILS, 2023); lower subjective technical and operational competencies in digital media use (YSL, 2024). | Unequal access to high-quality AI tools; potential exclusion from AI-based learning innovations; difficulties in meaningfully appropriating and navigating rapidly evolving technologies | Strengthen AI literacy across all schools with particular attention to learners from less advantaged backgrounds; integrate AI-related examples into existing BEE SECURE workshops and classroom materials; develop low-threshold, mobile-friendly online resources that can also be used independently at home; systematically signpost Helpline services. |
| Multilingual learners / children from non-Luxembourgish or non-German-speaking homes | Linguistic barriers in formal education; parents less able to assist with schoolwork. | AI can bridge language gaps (translation, explanation) but uncritical use may cause misunderstanding or bias. | Create multilingual BEE SECURE resources and webinars explaining safe and critical AI use; translate or subtitle key parent trainings (with AI); include examples of cultural bias and translation errors in awareness sessions; promote inclusive AI literacy in collaboration with schools, Maisons Relais and youth centres. |
| Children without effective parental learning support | Lack of educational guidance and digital supervision at home. | Over-reliance on AI tutors; exposure to misinformation or biased feedback. | Offer targeted BEE SECURE parent webinars on AI and online learning; produce easy-to-understand guides for families with low digital literacy; cooperate with schools to provide supervised after-school activities where safe AI use can be practised; highlight Helpline/KJT as a support option. |
| Girls and young women | Higher exposure to body-shaming, image-based harassment and online shaming. | Deepfakes, “nudifying” apps, synthetic sextortion and AI-assisted grooming intensify gendered digital violence. | Expand BEE SECURE’s gender-sensitive campaigns addressing AI-manipulated images and consent; include case-based modules on AI-generated sexualised content and digital boundaries in youth workshops; systematically communicate Helpline/Stopline services for victims of image-based violence. |
| Children and youth with a migrant background | Greater likelihood of harassment related to nationality, ethnicity or religion. | Algorithmic bias and AI-driven hate-speech amplification reinforce discrimination. | Strengthen cooperation with community organisations and migrant associations; include algorithmic bias and online discrimination topics in BEE SECURE trainings; develop multilingual campaigns promoting counter-speech and inclusive online culture; refer to Helpline/Stopline for reporting hate content. |
| Adolescents with low digital resilience or mental-health vulnerabilities | Psychosomatic complaints, loneliness, low self-esteem; higher susceptibility to manipulation. | AI-enhanced harassment (voice cloning, impersonation, automated shaming) exacerbates distress. | Integrate AI-related online safety into BEE SECURE’s mental-health and well-being workshops; train SEPAS and youth workers on recognising AI-mediated harassment; create resources on coping with synthetic bullying; ensure clear referral pathways to KJT/Helpline are communicated. |
| Younger users (early adolescents 11–15) | Limited cognitive maturity and risk awareness; rapid growth of cyberbullying in this age group. | AI-powered chatbots, deceptive games, and synthetic social interactions blur real-vs-fake boundaries. | Include early-age modules on “real vs synthetic” online content in BEE SECURE school and non-formal trainings (e.g. updates to Captain Kara, Superhelden, Ein Sprung ins Netz); develop child-friendly materials explaining AI-generated risks (deepfakes, filters, chatbots); provide parents with guidance on recognising manipulative chatbots and synthetic content. |
| Learners from socio-economically disadvantaged backgrounds | Limited access to devices and guided digital learning; lower CIL competences (ICILS, 2023); lower subjective technical and operational competencies in digital media use (YSL, 2024). | Unequal access to high-quality AI tools; potential exclusion from AI-based learning innovations; difficulties in meaningfully appropriating and navigating rapidly evolving technologies | Strengthen AI literacy across all schools with particular attention to learners from less advantaged backgrounds; integrate AI-related examples into existing BEE SECURE workshops and classroom materials; develop low-threshold, mobile-friendly online resources that can also be used independently at home; systematically signpost Helpline services. |
The overview in Table 2 shows that vulnerability to AI-related and digital risks in Luxembourg is shaped by intersecting socio-economic, linguistic, gender-related and psychological factors. Children from less advantaged or low-support households, girls and young women, migrant youth and emotionally fragile or younger adolescents (around 12 to 14 years old) are particularly exposed to educational and safety risks intensified by AI-driven technologies. These patterns align with insights from the internal staff survey, in which advisors reported that cases involving deepfake-based sexualised harassment and “nudifying apps” disproportionately affect girls, young women, and emotionally vulnerable adolescents. The survey findings therefore reinforce the relevance of these groups in the context of AI-mediated sexualised violence.