Safe AI for children - the most important answers

HillcrownAI is the European Safe AI Infrastructure for the children's world. We run own servers in Germany, an own moderation pipeline and an own CMS, and provide brands, publishers, schools, museums, sports clubs, platforms and public institutions with the building blocks they need to launch AI experiences for children without compliance or pedagogical risk. KinderGPT is the strongest use case of the platform today; further solutions follow through the /loesungen section.

Three components: a setup fee for onboarding and integration, an ongoing platform fee for the HillcrownAI platform, and a share of app revenue in which partners with their own IP or reach participate as licensors. Transparent components, no black-box list price.

Yes. HillcrownAI hosts exclusively in Germany, with no third-country transfer by default. We align with the EU AI Act, GDPR, German Youth Protection Law (JuSchG) and the Digital Services Act (DSA), and provide procurement-ready documentation on request (DPA, technical and organisational measures, AI Act mapping, JuSchG risk assessment). We carry the compliance load across all stages: GPAI obligations since 02 August 2025, standard obligations from 02 August 2026, health high-risk from 02 August 2027, education high-risk from 02 December 2027 – designed to meet every stage from day one.

KinderGPT is not a companion and not an engagement engine. There is no simulated empathy, no long-term memory, no para-social bond. Instead, parents are structurally informed, content is editorially cleared, and the system is designed for completed learning - not for daily active users. The architecture is the explicit antithesis of engagement-driven companion AI.

Content, IP, brand voices and characters of our partners remain the property of our partners. HillcrownAI only uses this content within the curated RAG pipeline. No model training occurs on partner content, neither by HillcrownAI nor by third parties. At the end of the contract, partners receive a complete content export and a documented platform teardown. The platform separates three layers of data sovereignty: user data sovereignty (conversations, learning histories), customer data sovereignty (partner IP, configurations) and infrastructure data sovereignty (DACH hosting, in-house moderation, no foreign cloud provider).

View data-sovereignty layers on the Platform page →

No. Data from KinderGPT and partner deployments is not used for model training. We use open-source base models, fine-tuned in-house, plus our own moderation models. External training pipelines have no access to children's or parents' data.

Parents get their own dashboard with activity overview, topic settings and escalation flags. Safety-relevant events (e.g. detected crisis topics) trigger structured parent notifications. Parents are informed - not replaced. The detail level is age-appropriate (4–14) and follows our klicksafe-aligned risk matrix.

We serve six use cases in a child's day: daycare & school, homework, club, family, weekend outing, content & idols. Eight solutions sit on the shared platform: KinderGPT, AI Tutor, Guide 3.0, Scan & Play, Event Companion, AI Ambassador, Digital Idol and Longvita (vertical for health/longevity coaching). More under /en/solutions and /en/worlds.

Agencies, systems integrators and publishers are certified in two days: day one architecture and compliance, day two hands-on with the CMS and moderation pipeline. Certified partners get a clearly defined margin model, co-selling support and access to Studios and Labs capacity. Apply by email to partner@hillcrownai.com.

We are a consortium partner in the ProFiZ project on protective-by-default AI for children and work with Rheinische Hochschule Köln, the University of Graz, the Catholic University of Eichstätt-Ingolstadt and the University of Oulu / Tampere University on pedagogy, developmental psychology, AI literacy and ethics reviews. A peer-reviewed usability evaluation of KinderGPT was published in 2025 by Springer Nature in the proceedings of EC-TEL 2025 (DOI 10.1007/978-3-032-03873-9_27, authors Pape and Zabolotna). A methodological follow-up paper from the ProFiZ context is in preparation. Publication list on request.

HillcrownAI is founded by Oliver Mauckner (Founder) and Tobias Friderich (Co-Founder), with roots in information technology, digital products, engineering and AI research. The sparring circle and partners come from research & pedagogy (ProFiZ consortium) and industry (senior operators in strategy, growth and finance).

Across our worlds we are running pilot conversations and partnerships with, among others, pilot schools across Germany, selected pilot municipalities and tourism regions, and selected publishers, broadcasters and FMCG brands. Concrete references are under NDA - we share them in conversation.

For brands, publishers, platforms and public bodies the fastest path is a 30-minute conversation with the founding team. For parents, KinderGPT launches separately as a consumer product. For agencies, the path goes through the Partner Program. For investors and journalists we recommend direct contact via the contact form or press@hillcrownai.com.

Architecturally and procedurally, yes. We are voluntarily highly regulated so our partners don't have to carry that burden alone. ISO/IEC 42001 certification is in preparation. Our data centre operator is ISO 27001 certified. AI-Act-compliant across all stages: GPAI since 02 August 2025, standard obligations from 02 August 2026, health high-risk from 02 August 2027, education high-risk from 02 December 2027 – we are ready.

Eighteen months of platform work have produced a turn-key solution. HillcrownAI Studios takes over onboarding, content ingestion and operations. HillcrownAI Labs covers API and co-engineering. Existing contracts become more valuable through this – no price shifts, no broken contracts.