AI research and consultancy from Germany, and the company behind c:node.
We work on one question: how do you build AI that has to justify its answers? For public administrations, banks and the German Mittelstand, "the model said so" is not an answer. Every statement needs a source someone can check, every decision a record, and nothing should leave the system without a person approving it.
c:node is a sovereign AI workspace for organisations. It connects a team's documents, mail and business systems with the c:node Graph — a continuously growing knowledge graph of EU and German law, public registers and market data — so that answers are grounded in sources rather than in model memory.
- Evidence, not eloquence. Answers cite the norm, register entry or document they rest on. 99.9%+ of the concepts in the c:node Graph are anchored to an official source; the share is measured on the graph itself, before any question is asked.
- Approval before effect. Mails, tasks and changes in connected systems are proposals until someone approves them.
- EU-first. Hosted in the EU, built for GDPR and the EU AI Act, and deployable on-premises down to air-gapped setups on open-weights models.
Explore the graph in the public observatory at graph.c-node.ai.
We publish methods, not implementations: how we evaluate, govern and grow systems that answer with evidence, mapped to public standards so others can apply and check them.
- grounded-ai-evals — a four-layer method for evaluating AI systems that answer with evidence: graph correctness, ontology hygiene, answer groundedness and baseline comparison, plus a release gate that blocks on named checks instead of a weighted mean. DOI 10.5281/zenodo.24100702
- governed-graph-growth — a technical note on keeping a continuously ingested, provenance-first knowledge graph trustworthy as it grows: source authority tiers, identity before enrichment, rule-classed linking, schema re-validation and growth health checks. DOI 10.5281/zenodo.24099664
- Product and pilots: c-node.ai
- Company: creativate.tech · contact@creativate.tech
- Research: Leonardo Bornhäußer · ORCID 0009-0005-6464-5179