Trust
Transparency
Version 1.0 · Last updated 2026-05-27
EU AI Act assumes you can show why the agent did what it did. So does any internal review when something looks off. Transparency is the difference between explaining decisions and apologizing for them.
Key concepts
Customer-facing transparency
The agent declares it is an AI on first contact. Disclaimer text is configurable per surface. Customers always know they are talking to an agent, not a human.
Decision traceability
Every agent answer carries the Sources it pulled from. Click an answer in Conversations, see exactly which Source was used.
Action traceability
Every Procedure run leaves a record. Who triggered it, what variables were used, what happened next.
Model and reasoning logs
Unless logs the model used, the prompt structure, and the retrieval context for each decision. Available to your DPO for review.
EU AI Act readiness
Unless is the Provider of the AI system under the EU AI Act. The platform is configurable for AI Act obligations: transparency notices to the customer, audit trails per decision, and the documentation a Provider needs to maintain. Most customer-agent deployments classify as limited-risk under the Act. Your account manager and the Compliance Center cover the formal classification path for your use case.
What you can do here
- Configure the AI disclaimer customers see per surface
- Pull a per-decision audit trail for any agent action
- Set reviewer access for your DPO or internal auditor
- Export transparency reports for regulator inquiries
- Reference the Compliance Center for the formal Provider documentation
When to use it
- During pre-launch compliance review for a new Moment
- When the DPO asks for an audit trail
- When a customer asks why the agent said something specific
- When preparing for an EU AI Act readiness review
How it works
The Engine logs the inputs, the model used, the Sources retrieved, and the output for every decision. Logs feed Conversations, Accountability, and the per-decision audit trail your DPO can read.
Frequently asked questions
How do I change the AI disclaimer customers see?
Open Trust > Transparency and edit the disclaimer text. Pick the surface (marketing site, product, helpdesk). Save. The disclaimer fires on first contact in every session by default.
How do I classify my agent under the EU AI Act?
Talk to your account manager. Most customer-agent deployments are limited-risk. The formal classification documentation lives in the Compliance Center, with the rest of the Provider materials.
Where is the audit trail?
Open any conversation in Conversations and view the decision trail panel. For a workspace-wide export, use the audit export in Trust.
How do I give my DPO access without giving them full admin?
Create a DPO role in Team and roles with read access to Trust and Conversations only.
What is logged for each decision?
Inputs (customer message, Living Context state), Sources retrieved, model used, output text, Procedures triggered. Plus the timestamp and user identity (where applicable).
Does the agent always say it is an AI?
Yes. The disclaimer fires on first contact in every session by default. You can change the wording but not turn it off.
