Passport
Defines identity, owner, purpose, scope, lifecycle state, and baseline evidence needs.
Evidence & Accountability Model
Preserve the record chain behind AI worker review, control decisions, and later reconstruction.
Scaled Agents treats evidence as a set of linked operating records: Passport scope, Human Review, Toll Gate decisions, Runtime Permit posture, workflow events, and Stamp or audit-export outputs.
This page is public context for how accountability works. It does not expose customer evidence, secrets, reviewer notes, raw logs, or a completed audit packet.
Context
Evidence should explain who decided what, when, and under which boundary.
Auditors, security reviewers, and internal governance teams usually need to reconstruct decision paths, not just read a policy statement.
Scaled Agents is designed to keep that reconstruction grounded in linked records rather than isolated screenshots or untraceable narrative claims.
- Linked records instead of isolated assertions
- Customer-safe summaries instead of public disclosure of sensitive evidence
- Human accountability remains visible at each review step
Record Chain
These records create the accountability model.
Human Review
Captures accountable reviewer decisions, escalation, remediation, and exception handling.
Toll Gate & Permit
Shows whether a request was allowed, blocked, paused, or routed based on current authority.
Stamp & Export
Preserves traceable event markers and customer-safe summaries for later review.
Claim boundary This model helps explain how evidence can be organized. It does not guarantee that a customer has completed every required test, review, or independent assessment for a deployed environment.
Evidence Views
Different audiences should see different levels of detail.
High-level descriptions of the record model, control boundaries, and customer-managed accountability.
Authorized Passport details, evidence references, approvals, exceptions, and environment-specific decision history.
Customer-approved scope, selected evidence extracts, test outputs, and the separation-of-duties trail required for a specific engagement.