See their own submission
Can access published programme information and their chosen reporting pathway. No access to other employees’ cases.
Observe, not adjudicate
Can submit structured observations and see only authorised team-level climate outputs.
Context-rich case access
Receives the minimum necessary case detail, documents rationale and selects a proportionate route.
Role-gated escalation
Access only when the chosen pathway and organisational protocol justify it.
Thresholded intelligence
Sees aggregate trends and response health, not a leaderboard of “risky employees.”
Process accountability
Can inspect version, access events, review timing, overrides and disposition logic without altering cases.
Some capabilities should remain deliberately unavailable.
Employees are told they are participating in a trust and safety programme. Scoring mechanics may be protected against gaming, but the purpose is not hidden.
Sentinel does not use webcam expressions, voice stress, micro-expressions or similar proxies to infer mental state.
Religion, ethnicity, nationality, political belief, disability and other sensitive traits do not become threat indicators.
Signal outputs cannot directly trigger termination, suspension, hiring rejection, promotion denial or discipline.
Performance must be defined against a specific outcome, population, scoring version and evaluation window.
Every consequential review should leave a trace.
Authenticated enterprise workspaces now use a server-side audit log; the public demonstration falls back to local browser history.
No review actions recorded in this browser yet.