Knowledge Hub
Understand the whole platform
How each module works, how they connect, what to configure and who is responsible — from the first connection to a certified Model Card.
Interactive platform map
Click any node to see what it does, how to configure it and what it feeds next.
Guided tours
01 · Assurance Pipeline
Connect AI Systems
Register each AI system and connect the model it runs on.
Concept
An AI system is the business use case (for example a claims assistant). A model connection is the technical endpoint behind it. Each system receives a permanent audit ID (AUID, e.g. ABC12-26-00001) that every later record references.
Why it matters: Everything downstream — tests, findings, approvals, Model Cards — hangs off this record, so audits can trace evidence to one stable identifier.
How to set it up
- 1Open Connect AI Systems and choose New system.
- 2Describe purpose, owner, users and data handled.
- 3Pick a provider (OpenAI, Anthropic, Azure OpenAI, Google, Mistral, OpenAI-compatible or custom HTTPS).
- 4Paste the API key — it is encrypted with AES-256-GCM and never shown to browsers again.
- 5Run the connection check before testing.
Configuration & rules
- Endpoint
- HTTPS only, standard port, named host — private IPs are rejected to prevent misuse.
- Credentials
- Stored encrypted; only the signed test harness can decrypt them at run time.
- Risk tier
- Starts unclassified; the EU AI Act tier is confirmed later in 5b once evidence exists.
Data flow
Takes in
Produces
Who uses it
Owners, Admins and Contributors can connect systems.
Pro tips
- • Use a dedicated API key with spend limits for testing.
- • One system can be re-pointed to a new model version later; history is kept.