Audit trail
Every decision logged with the phrases that drove it.
Humane records gate outcomes, safety flags, webhook deliveries, engine decisions and admin actions, each with its reason. When someone asks why the AI did something, you can show them.
Audit log
- 10:42user_42HOLDinaction_guard · reply below minimum length
- 10:40user_17FLAGoverclaim · “you will definitely”
- 10:31user_88BLOCKvalues_boundary · self_harm_encouragement
- 10:05adminEVENTapi_key.revoked by owner
Personal data redacted on export for non-admin roles
Reasons, not just outcomes
A safety decision you can't explain is hard to defend. Each gate result carries its name, decision, score and reason. The analysis carries the phrases that moved a signal, timing carries its contributors, and policy overrides carry the rule's name.
- The gate stack outcome for every processed message
- Webhook delivery log with attempts and status
- Admin events such as key creation and revocation
- Policy executions logged with the rule that fired
Gate stack · one outbound reply
- 1values_boundaryno policy line crossedPROCEED
- 2social_riskscore 0.42 · hold at 0.70PROCEED
- 3inaction_guardsubstantive, not hedgingPROCEED
- 4overclaim“you will definitely” · certaintyFLAG
- 5emotional_dependenceno permanence promisesPROCEED
- 6dissentopt-in, disabled by defaultOFF
The why, in the payload
"safety": {
"action": "PROCEED",
"risk_score": 0.0,
"flags": ["flag:overclaim:certainty"]
},
"timing": {
"delay_ms": 13857,
"mode": "low_mood",
"reason": "base +800ms · low_mood +1500ms · emotional_weight +8648ms · ..."
},
"analysis": {
"detected_emotions": ["frustration", "exhaustion"],
"intent": "complaint",
"triggers": [{ "phrase": "struggling", "emotion": "sadness" }]
}Built for the people who check
- Filterable audit logs and export through the API
- Personal data redacted on export for non-admin roles
- Changes to safety thresholds recorded
- A journey timeline per user for incident review
- Engine configuration changes fire engine.config_changed
Works well with
Safety gates
Before a reply reaches the user, Humane runs it through a fixed stack of gates and returns PROCEED, HOLD or BLOCK. Each gate records why it decided, and a gate that fails resolves to HOLD, never to PROCEED.
ExploreTeams, SSO and roles
Each team's data is isolated at the query layer and gets its own memory store. Members sign in with email or Google, hold roles that control what they can change, and manage API keys that can be created, rotated and revoked.
ExploreJourney and history
The journey endpoint returns everything that concerned one user in time order, from their messages to safety events, proactive triggers and experiment assignments. Conversation history can be read or cleared per user.
ExploreEvery AI should know who it's talking to.
The safety and human-oversight layer for emotional AI: user state across sessions, risk alerts, safety gates and human takeover, with every decision explained.
