| Alert-only posture | Risk rules raise alerts for human screening; they do not auto-block funds. The default for every production rule today. |
| Auto-resolve | When the identity suite returns a soft-reject, the AI false-positive synthesiser may auto-override if confidence is above the configured threshold. |
| Case | The unit of work — a transaction, KYC, or KYB submission flowing through a workflow. |
| Decision Console | The per-case evidence view at /decisions/:id. |
| Decision Queue | The live work list at /queue. |
| Hard reject labels | Identity-suite labels that bypass auto-resolve and route to manual review with a deterministic-fail recommendation: forgery, sanctions hit, blocklisted, criminal, fraudulent patterns, deepfake. |
| Modular-by-client | A rule that has the same shape across clients but per-client thresholds. Every production risk rule supports it. |
| Resold vendor | A provider Frayme manages credentials for. The client sees the data, not the API key. |
| RFI | Request For Information — formal evidence request to the relationship client during alert investigation. |
| Soft reject labels | Identity-suite labels eligible for auto-resolve: photo blurred, document photo resolution, incomplete document, bad selfie, bad face matching, document page missing. |
| Two-layer region gate | The pattern of requiring (1) the country is on a commercial allowlist AND (2) not on a prohibitions list. |