SafeNutriKids
SafeNutriKids is an innovative AI-powered nutrition education platform designed for children aged 6-12. Addressing a critical gap in digital nutrition tools for younger audiences – and a wider inequity in digital literacy – the project combines personalised, gamified learning with a zero-trust architecture that protects sensitive user data by design. A dual-interface platform enables children to freely explore healthy eating while learning to navigate digital tools safely and critically, while parents are guided to model responsible data and consent management through their own dashboards.
Developed under the DRG4FOOD initiative, SafeNutriKids demonstrates how Digital Responsibility Goals (DRGs) can shape both pedagogy and platform design – from privacy-first data handling to modular, open-source localisation tools that align AI prompts and meal recommendations with region-specific foods, languages, and nutritional contexts, replacing one-size-fits-all dietary models with locally grounded data. By merging inclusive, child-centred education with verifiable digital responsibility, SafeNutriKids offers a transferable blueprint for innovators seeking to build trustworthy, culturally sensitive, AI-driven food-system solutions for children.
WHY THIS MATTERS
Children today learn and make choices in a digital environment that profoundly shapes what they eat, how they think about health, and whom they trust for information. Yet most nutrition apps remain designed for adults – focused on productivity, calorie tracking, or lifestyle branding – and rarely account for children’s learning needs, safety, or agency. Too often, young users are treated as passive recipients rather than active participants, while their personal data is collected with little transparency to fuel equally opaque business models.
This absence of genuinely child-centred design leaves families without the digital tools to help children build essential life skills – digital literacy, healthy eating, and critical thinking – while upholding ethical and regulatory standards for minors’ data. SafeNutriKids responds to this gap by embedding the requirements of the EU GDPR and the emerging AI Act directly into its design. Privacy-by-default settings, parental dashboards, and explainable AI functions translate these ethical and regulatory expectations into everyday learning: children and parents experience consent, transparency, and data control as part of the educational process itself.
By gamifying healthy eating, SafeNutriKids aims to turn a lifelong skill, understanding nutrition, into an engaging, self-directed experience that builds both confidence and critical awareness. It also ensures that personal and nutritional data, among the most sensitive categories of information, remain protected and never repurposed for marketing or profiling.
Importantly, SafeNutriKids also addresses the cultural and ethical dimensions of digital learning. Its AI models are trained and validated with diverse regional data to prevent bias and reflect local food traditions, languages, and values. An approach that makes the platform not only safer, but more relatable – recognising that food is both deeply personal and culturally specific. In doing so, SafeNutriKids reframes digital nutrition education as a matter of public responsibility, one where inclusion, fairness, and data protection are inseparable foundations of a trustworthy and equitable digital food system.
WHAT THE SOLUTION DOES
The SafeNutriKids project transforms the everyday challenge of teaching healthy eating into a shared digital experience for children and parents.
Imagine a nine-year-old opening the app after school. Instead of another lesson or lecture, they enter an animated, game-like world where nutrition becomes an interactive adventure. Each module revolves around familiar foods – the ones found at home or in local shops – and adapts to regional dietary traditions. Through storytelling, character-based navigation, and reward-based learning tasks, children explore how different foods contribute to energy, growth, and well-being. As they play, the app’s AI adapts each child’s learning path to their pace, preferences, and cultural context, suggesting locally familiar foods and new challenges that keep curiosity alive.
Meanwhile, a parent or caregiver accesses the companion dashboard – a private space to follow progress, explore child-friendly meal suggestions, and understand what their child is learning. The AI assistant, powered by ChatGPT and validated by nutrition experts, recommends meals based on age, allergies, and local food availability. Built on a zero-trust, privacy-by-design foundation, the system combines a Unity-based frontend with a Firebase and Google Cloud backend for real-time updates and GDPR-compliant data handling. Privacy dashboards allow families to review, modify, or delete information at any time, while transparency layers show how recommendations are generated. In this way, both children and adults develop a practical understanding of how AI works – transforming algorithmic transparency into everyday digital literacy and trust.
Pilots in Estonia, Bulgaria, and Turkey confirmed the platform’s appeal and adaptability. Children played with the app as a daily routine rather than a school assignment, and the gamified design sustained engagement while culturally localised content made lessons more personal, recognisable and relevant. Parents reported stronger awareness of data responsibility and a clearer sense of how technology can complement, rather than replace, their role in nutrition education.
Beyond nutrition, the SafeNutriKids model demonstrates how AI, privacy-by-design, and culturally adaptive content can work together to create more trustworthy, inclusive learning environments, a transferable template for other fields where digital literacy, trust, and behavioural change go hand in hand.
HOW YOU CAN USE IT
SafeNutriKids offers different entry points for various stakeholders:
For families and communities: SafeNutriKids is designed as a shared learning experience for children and their families. At home, it turns conversations about food and technology into playful, everyday moments. Children explore nutrition independently through games and storytelling, while the companion dashboard helps families of all shapes to stay involved, offering a simple way to follow learning progress, adjust preferences, and understand how data and consent are handled. In workshops and pilot trials across Estonia, Bulgaria, and Turkey, many families reported that the app created new opportunities to talk about food traditions and healthy eating together, blending play, culture, and learning. By framing food as both a digital and social experience, SafeNutriKids nurtures curiosity, trust, and shared responsibility within the household and beyond.
For educators and schools: Beyond home use, pilots demonstrated the app’s potential for integration into classroom and extracurricular programmes. SafeNutriKids offers an engaging digital tool that supports nutrition and digital literacy education, complementing national health curricula and initiatives. Teachers involved in the pilots in Estonia, Bulgaria, and Turkey used the app to facilitate group discussions and activities around healthy eating and responsible technology use. Aggregated dashboards enabled them to observe learning progress and engagement levels without collecting personal data. Workshops and feedback sessions confirmed that SafeNutriKids can be easily incorporated into teaching practice without specialist training, providing high engagement and inclusivity across cultural contexts.
For developers and innovators: SafeNutriKids contributes reusable components that demonstrate how to implement privacy-first, child-safe AI in practice. The platform’s architecture – combining a Unity frontend, Firebase and Google Cloud backend, and explainable AI module – is fully documented, with open-source localisation templates that can be adapted to other learning domains. Developers can build on its zero-trust approach, privacy dashboards, and cultural adaptation methods to design digital experiences that are transparent, inclusive, and regulation-ready under the GDPR and AI Act.
For policymakers and health authorities: SafeNutriKids offers a concrete example of how Digital Responsibility Goals (DRGs) can inform both educational and public-health policy. The project demonstrates that responsible AI and data governance can coexist with child-centred learning and measurable health outcomes. Its documentation details how privacy, consent, algorithmic transparency, and bias mitigation were embedded from concept to deployment, providing a model for future standards and certification frameworks. For policymakers, SafeNutriKids illustrates how GDPR and forthcoming AI Act requirements can be operationalised in practice – turning regulation into design guidance rather than compliance overhead. For health authorities, it presents an adaptable tool for advancing national nutrition strategies, digital literacy goals, and safe data use in family and school environments. Together, these outputs provide an actionable blueprint for governing trustworthy digital learning and nutrition initiatives in a way that protects children’s rights while strengthening public trust in technology.
Together, these outputs make SafeNutriKids more than a single application – it is a shared resource for teaching, designing, and regulating digital responsibility in practice.
DIGITAL RESPONSIBILITY IN PRACTICE
SafeNutriKids embeds all seven DRGs throughout its design:
Digital literacy is not only a learning outcome but also a design principle in SafeNutriKids. The pedagogical approach, developed with educators and child-psychology experts, follows a learning-through-play model validated by Trakia University to ensure age-appropriate engagement for children aged 6-12. Storytelling, character-based navigation, and reward-based tasks allow children to make choices, explore feedback, and build confidence in a protected environment. Simple icons, clear audio cues, and colour-coded guidance support different literacy levels and ensure accessibility for younger users. Parental dashboards explain how AI recommendations are generated, turning transparency into a shared form of algorithmic literacy for both adults and children.
Privacy is built into SafeNutriKids from first use. During onboarding, families choose what data to share through modular consent options that cover AI-assisted recommendations, dietary tracking, and learning analytics. Parents can later review, modify, or withdraw these choices at any time through a dedicated privacy dashboard. When an account is deleted, identifiable information is erased, with only aggregated, anonymised statistics retained for evaluation. No data is ever shared or reused for marketing or profiling. By embedding GDPR principles; consent, minimisation, transparency, and the right to erasure-directly into user flows, the project turns privacy from a legal checkbox into a daily, understandable practice for both children and adults.
Data fairness is a core value in SafeNutriKids and a foundation of trust. It ensures that the data and algorithms behind nutrition learning reflect real dietary diversity rather than abstract averages. The platform’s localisation layer adapts content and AI prompts to region-specific foods, languages, and traditions, while nutrition experts in Estonia, Bulgaria, and Turkey validated each module for accuracy and inclusivity. Personalised recommendations respect declared allergies and preferences, and manual reviews of AI-model outputs remove culturally or gender-biased examples. By replacing one-size-fits-all, Western-centric diet models with locally grounded guidance, SafeNutriKids shows how fairness in data design translates into genuine equity and trust in digital learning.
SafeNutriKids treats algorithmic transparency and human oversight as inseparable foundations of trust. All AI interactions occur server-side, with user identifiers stripped before processing and all requests encrypted in transit. Prompts are written to favour clarity, avoid sensitive or exclusionary language, and include short explanations or source cues in responses. Every AI output is stored in anonymised logs that support ongoing human review: nutrition and education experts regularly check model responses for factual accuracy, tone, and cultural balance, providing feedback that guides prompt and content updates. These technical safeguards and human-in-the-loop mechanisms work closely together to ensure that the AI remains explainable, auditable, and aligned with SafeNutriKids’ educational values – demonstrating how trustworthy algorithms can serve children safely in real learning environments.
Transparency in SafeNutriKids is both practical and participatory. Within the app, icons clearly mark AI-generated content, and each recommendation includes a short explanation such as “based on your recent choices” or “using locally available foods.” The parental dashboard provides real-time visibility into what data is stored and how it is used, while also allowing families to adjust privacy settings and consent preferences directly. When data sharing options are changed, the dashboard immediately updates the scope of AI recommendations – making the link between consent and output tangible and easy to understand.
Beyond the interface, transparency extends to architecture and governance. All AI interactions are logged in anonymised form and can be reviewed by authorised experts, while the project’s API specifications and consent-management modules are publicly shared through the DRG4FOOD Toolbox.
SafeNutriKids reinforces human agency by keeping children’s learning exploratory and self-directed within a safeguarded framework. The app encourages decision-making through play and reflection rather than instruction, allowing children to build confidence and curiosity while staying protected. Parents retain visibility and consent control through the dashboard, guiding rather than policing their child’s activity.
No behavioural nudging, tracking, or health-linked rewards are used, and all AI suggestions remain advisory and optional. Personal data are pseudonymised, and attributes such as age or gender are applied only to adapt educational content –not for profiling or segmentation. In this way, SafeNutriKids ensures that technology enhances, rather than replaces, human judgement and that every learner’s dignity and identity remain respected.
CONTRIBUTION TO THE TOOLBOX
The SafeNutriKids project contributes a set of practical, reusable enablers that support culturally sensitive, multilingual, and child-appropriate digital nutrition education. The components include a fully open-source localisation layer, reference data tables, example catalogues, and methodological guidance. Additional modules relating to parental access are provided as structured guidelines with open-source developments to be further contributed.
- Multilingual localisation framework
The localisation layer implements a metadata-based approach to cultural and linguistic adaptation of nutrition content. Rather than translating fixed text, the system tags each content variant with structured metadata-such as language, country, cultural context, dietary relevance, and age range-and selects the best match at runtime. This ensures that children receive examples and recommendations that fit their cultural and linguistic context while maintaining educational consistency. The DRG4FOOD Toolbox package includes the localisation engine, schemas, reference tables, and worked examples, enabling developers to integrate transparent, configurable, and culturally sensitive localisation into their own applications.
View on DRG4FOOD GitHub
- Parental consent and data-access module
A reusable framework for managing parental consent, child data access, and GDPR-compliant control flows in digital learning environments. The module defines how platforms should authenticate parents, verify consent before activating child-facing features, and provide transparent mechanisms for reviewing, modifying, or deleting data. It emphasises auditability, child-friendly communication, and responsible handling of personal information in alignment with GDPR Article 8.
The contribution currently includes the design-level documentation, outlining recommended workflows, data-handling practices, consent states, and integration patterns for secure implementation in child-focused applications. The technical enabler (code implementation) is pending.
View on DRG4FOOD GitHub
Together, these contributions form a robust reference for building responsible AI in child-targeted food education tools.
IMPACT AND OUTLOOK
The SafeNutriKids project has demonstrated that responsible, well-designed technology can make nutrition and digital literacy learning engaging and effective for children and families. Across pilots in Estonia, Bulgaria, and Turkey, more than 250 participants – including 145 children and 112 parents –tested the app in home and school settings. Module completion exceeded 90%, and parental satisfaction averaged 4.3 out of 5, validating both engagement and usability. The results confirmed the platform’s cultural adaptability and readiness for scale, with education authorities and school partners in all three countries expressing interest in continued use and further deployment.
Beyond educational impact, the project established a replicable governance and business model for child-focused digital tools. The consortium is exploring a hybrid sustainability strategy combining public-sector partnerships and family subscriptions, aligned with national priorities: integration into e-education frameworks in Estonia, public-health initiatives in Bulgaria, and collaborations with the Ministry of National Education in Turkey.
Looking forward, SafeNutriKids will expand its reach across Europe, supported by collaborations with ministries, parent associations, and educators. Planned next steps include advancing AI explainability, strengthening health-system integration, and creating teacher training programmes on digital responsibility in nutrition education.
By prioritising children’s rights, privacy, and well-being alongside educational effectiveness, SafeNutriKids provides a practical model for trustworthy, values-based AI in the food and education systems – showing how the DRG4FOOD principles can scale from pilot innovation to public impact.
QUICK FACTS
- Funding
- DRG4FOOD Open Call #1
- Use case
- Targeted Nutrition
- Partners
- Start date
- Apr 2024
- End date
- Apr 2025
- Resources