How do we prove to an auditor what the AI was told?

Auditors rarely ask what a system answered. They ask what it was working from — which instructions were in force, which records were retrieved, who the requester was, and whether that person was entitled to the material that informed the response. Those are different artefacts from a transcript of outputs, and most deployments cannot produce them, because the context window is assembled in memory at request time and discarded once the answer returns.

Centralpoint retains the assembly rather than only the result. The AI Interaction Log captures each execution with the skills that loaded, the prompt that governed it and the token accounting for that request; Dialogue History preserves the conversation those executions belonged to. Because skills and prompts are records in the organization's own SQL environment rather than settings inside a vendor's product, they carry version history — so the instruction set as it stood on the date in question can be reconstructed rather than approximated. The Prompts and Skills registries expose current state as live feeds, which lets an auditor see what governs the system now and what governed it then, from one source.