Definition
The lag between harm occurring and the system detecting or responding to it. Automation can scale impact faster than oversight can respond, so shrinking the lag takes friction that slows fast automated steps and automatic halts.
Tests that show it is in place
- Harm is detected before the system can scale it.
- Harm alerts are tied to halt, rollback, or escalation controls.
Hypothetical example
- A clinic triage chatbot keeps routing patients while harm reports queue for hours, delaying escalation.
References