Mandatory Deployment Safety Guidelines
The structural prerequisites required to pre-qualify for Agentic E&O Insurance. We protect the insurance pool by enforcing algorithmic safety.
Isolated Prompt Execution
Agents must execute state-changing actions in isolated, sandboxed environments with strict parameter validation before API execution.
Immutable Telemetry Logging
Every decision node must stream logs directly to the aaieo "Orange Box" via read-only API. Local, alterable logs will void coverage.
Human-in-the-Loop Thresholds
Financial transactions or legal communications exceeding the designated Risk Threshold ($1,000+) must require an Accredited Human sign-off.
Continuous Hallucination Monitoring
Models must run parallel sanity checks against deterministic rule-engines. Output that violates deterministic logic must be flagged and halted.
The Profitability Contingency
If an incident occurs and the telemetry proves these guidelines were followed, the insurance triggers an automatic payout.
If the telemetry proves these guidelines were bypassed, the claim is legally denied.