GERMAN AI EVIDENCE NETWORK

Where public AI data actually connects.

Projects, organisations, systems, procurements, repositories and budget titles remain separate entities. The network shows only published relationships—with evidence, a join rule and visible gaps.

stable identifiersone hop, not graph chaosevidence on every edgeno generative AI

3,626searchable entities

2,838evidenced connections

7source snapshots

3visible linkage gaps

THE DECISIVE BOUNDARY

Linked means evidenced—not complete.

An edge says only what its published identifier, source statement or declared mapping supports. Similar names never create identity; missing edges do not prove absence.

exact · source-reported · declared

Search the evidence network

Entity type

COMPACT ONE-HOP GRAPH

One centre. Its evidenced neighbours.

Select an entity. The view shows at most eight directly evidenced neighbours, prioritised by evidence class and connection count. Select an edge to inspect its proof.

Selected centreFRAUNHOFER GESELLSCHAFT ZUR FORDERUNG DER ANGEWANDTEN FORSCHUNG EVActor · actor:cordis-pic:999984059

Open dossier

Further direct neighbours · 115

3,626 results

EVIDENCE BEFORE NARRATIVE

One shared layer without blurring the sources

The network normalises presentation and navigation. Meaning, money type and claim boundary remain attached to every source dataset.

Separate entities

Organisation, project, procurement, system, repository and budget title remain distinct types with their own stable identifiers.

Classify edges

Exact identifiers, source-reported relationships and declared mappings remain visibly separate.

Retain evidence

Every edge resolves to a source field, snapshot or original URL. Names alone never create a link.

Publish gaps

Unsupported transitions remain visible as Missing Bridges instead of being filled by a model or similarity.

Source snapshots used

german-ai-actor-registryactor-registry · retrieved 21 July 2026

german-public-ai-outcome-evidence-ledgeroutcome-evidence-ledger · retrieved 21 July 2026

german-public-ai-system-evidence-passsystem-pass · retrieved 21 July 2026

german-public-sector-ai-code-radarcode-radar · retrieved 21 July 2026

german-public-sector-ai-dependency-atlasdependency-atlas · retrieved 21 July 2026

procurement-chronicleprocurement-chronicle · retrieved 21 July 2026

german-federal-ai-budget-monitorbudget-monitor · retrieved 20 July 2026

Clearly explained

Does every similar name become one entity?

No. Names are display fields, never sufficient join keys. Entities merge only where a stable identifier, an explicit source relationship or a published declared mapping supports the connection.

Is this a money-flow graph?

No. Budget appropriations, grants, procurement values, payments and actual expenditure remain separate facts. The network connects them only when an official identifier explicitly supports that relationship.

Why does the graph show only one hop?

A compact neighbourhood keeps every visible edge inspectable. A global hairball would conceal both evidence quality and missing links.

What is a Missing Bridge?

It is a documented linkage gap: two domains could answer a useful question, but the bounded public sources do not publish a reliable identifier or relationship. It is not an inferred edge.

Does the network use generative AI to create links?

No. Collection, identifier matching, edge classification and publication are deterministic. Unsupported similarities stay unlinked.

Need a defensible evidence graph for your domain?

We turn fragmented public records into inspectable products with stable identities, explicit claim boundaries and permanent source paths.

Discuss a data project Explore all tools