Question 1

Does Digicane train foundation models on our data?

Accepted Answer

By default, no. We design retrieval and logging so your documents are used to answer — not to retrain public foundation models — unless a separate agreement says otherwise.

Question 2

How do you reduce hallucinations on company facts?

Accepted Answer

We use RAG with citations, constrain tools, and escalate when retrieval confidence is low.

Question 3

Can regulated actions run without a human?

Accepted Answer

They can be configured that way, but Digicane defaults sensitive steps to human override until you explicitly widen automation.

Question 4

Where can prompts and transcripts be stored?

Accepted Answer

Storage location follows your residency and retention choices — including India-based options.

Question 5

Do you support DPA language?

Accepted Answer

Yes — commercial paperwork can include data processing terms aligned with your counsel’s requirements.

Question 6

How are AI incidents reviewed?

Accepted Answer

Audit logs, eval failures, and override rejects feed into change control and runbook updates.