HillcrownAI

Infrastructure

Five building blocks of a sovereign Kids-AI stack

HillcrownAI runs on five independent infrastructure pillars: German servers, open-source models, in-house moderation, multiguard architecture and an in-house CMS. Each pillar is documented and auditable.

Read the safety matrix

Why this matters

Sovereignty is not a marketing claim

European children's data should not flow into US-based foundation models without controls. We rebuilt the stack so that every byte sits in a known datacentre, every model can be audited, and every escalation path stays inside the EU.

Five building blocks

What "in-house" actually means

01

DE servers

Workloads run in ISO 27001-certified datacentres in Germany. No US transfers by default; transfers only on explicit, documented request and with SCCs in place.

02

Open-source base models

We build on open-source foundation models. No black box, no surprise licensing changes, no one-vendor lock-in for our partners.

03

In-house moderation model

Our moderation model is trained and operated by HillcrownAI. Decisions are explainable, escalations have a human owner.

04

Multiguard architecture

Every response passes through dataset curation, in-house moderation and runtime guardrails. Defence-in-depth, not a single content filter.

05

In-house CMS

Editorial teams, partners and parents work in our own CMS. No third-party content workflows that leak children's data into adjacent products.

Compliance

Aligned with EU regulation

Every infrastructure decision is mapped to a specific obligation. The most relevant frameworks today:

Read on

Related pages on trust & compliance

Architecture, safety and research belong together — the three pages of the trust lane each deepen one aspect.

Architecture deck on request

We share an architecture deck with diagrams, datacentre locations and our audit posture under NDA. Tell us about your evaluation and we'll set up a 45-minute call with the engineering lead.