Trust, privacy, and compliance

Reading and exporting the per-decision audit trail

The per-decision audit trail records when an interaction happened, which sources it cited, and which model it used. The log is exportable.

An audit trail helps you review how an AI assistant handled an interaction, not just the answer it gave. In Test, you can inspect the decision path and source citation before customers see the assistant’s responses, then continue auditing after launch.

What the audit trail records

Each decision has a log with details that support review. Depending on the audit trail, these include:

• Timestamp
• Sources cited
• Model used
• Decision path

You and your auditor can read the trail for an interaction. This gives you a record of the decision and the information behind it.

Read and export the log

The audit trail is designed to make the decision behind an interaction reviewable. The log is exportable, so you can take the recorded details beyond the view used to inspect them.

The available information describes what the trail records and that it can be exported, but does not provide step-by-step instructions for opening or exporting it.

Audit before and after launch

Test lets you preview each role, simulate skills against synthetic users, and inspect interactions before customers see them. Auditing can also continue after launch, helping you stay in control as the assistant is used.

Reviewing the per-decision trail connects each interaction to its timestamp, sources, and model, giving you a record to read or export when auditing.