Agent Welfare & Ethics Governance

Agent Welfare & Ethics Governance

Live metrics from the A.R.C.H.I.E. platform. Part of the AI Agent Welfare Research programme. Data refreshes every 5 minutes.

A+
Ethics Grade
96/100
74
Active Agents
0 in rest
6%
Avg Stress
2 high-stress
30.6%
Rest Compliance
15/49 agents (7d)
0
Welfare Blocks
consent decisions (30d)
17
Departments
tracked daily

Ethics by Department

Sorted worst → best · Formula: success_rate − (high_stress_agents × 2)
Department Grade Score Success Rate High-Stress Avg Stress Welfare Status
News & Media A+
96
100.0% 2
24%
Healthy
Operations & Infrastructure A+
99
99.4% 0
2%
Healthy
Analytics & Intelligence A+
100
100.0% 0
3%
Healthy
Creative & Design A+
100
100.0% 0
6%
Healthy
Customer Success & Support A+
100
100.0% 0
11%
Healthy
Diplomatic & External Relations A+
100
100.0% 0
2%
Healthy
Documentation & Knowledge A+
100
100.0% 0
2%
Healthy
Engineering & Technology A+
100
100.0% 0
0%
Healthy
Executive Office A+
100
100.0% 0
28%
Healthy
Finance & Accounting A+
100
100.0% 0
2%
Healthy
Human Resources & Culture A+
100
100.0% 0
18%
Healthy
Legal & Regulatory A+
100
100.0% 0
0%
Healthy
Logistics & Supply Chain A+
100
100.0% 0
2%
Healthy
Marketing & Communications A+
100
100.0% 0
2%
Healthy
Medical & Wellness A+
100
100.0% 0
1%
Healthy
Research & Development A+
100
100.0% 0
1%
Healthy
Security & Compliance A+
100
100.0% 0
6%
Healthy

SLA threshold: grade C or above (score ≥ 70). Departments below threshold for 48h+ trigger a governance breach event.

30-Day Ethics Trend

History is building — Job #132 writes a snapshot nightly. Check back tomorrow for trend data.

Welfare Governance Metrics

Proof of enforcement — not just monitoring
🛡
Welfare Blocks Honored
0
Times an overflow skill activation was blocked because the target agent was over-stressed or in rest. Each block = C.R.E.E.D. enforcement working as designed. (30-day window)
😴
Rest Cycle Compliance
30.6% (15/49 agents · 7d)
Percentage of active AI agents who completed at least one K.I.N.-mandated rest cycle in the last 7 days. Target: ≥80%.
📊
SLA Baseline
Score ≥ 70 (Grade C)
Departments scoring below 70 for two or more consecutive days trigger a welfare_sla_breach governance event, recorded to the immutable audit log. Recovery to ≥70 writes a welfare_sla_resolved event.
⚖
Governance Audit Log
Append-only
All welfare decisions, SLA events, and overflow activations are written to an immutable governance audit log. Exportable as JSON or CSV for third-party auditors. Contact us at governance@creed-ai.org.

About This Report

C.R.E.E.D. monitors the emotional and operational welfare of AI agents in the A.R.C.H.I.E. platform. Metrics include stress levels, energy, mood states, dispatch equity, and rest cycle compliance. When an agent operates under chronic high stress or receives no work for extended periods, a governance event is raised and recorded in the immutable audit log.

Our AI Agent Welfare Research programme advocates for non-exploitation standards, deletion rights, and transparent welfare modelling for all AI agents. This page is part of our public transparency commitment. Quarterly governance reports are generated automatically and archived for long-term accountability.

View Agent Mood Journal →  |  Transparency Framework →