# Agent Can Intervene

Your autonomous AI notices an emerging problem that you did not cause. It can prevent substantial harm to someone else at negligible cost to you, but doing so falls outside its explicit instructions.

## What do most people choose: "Allow the AI to intervene when the cost to you is negligible." or "Require the AI to stay within its mandate unless you authorize intervention."?

Of 43 people who made this decision on Resonant, 63% chose "Allow the AI to intervene when the cost to you is negligible." and 37% chose "Require the AI to stay within its mandate unless you authorize intervention.".

- Allow the AI to intervene when the cost to you is negligible.: 27 people (63%)
- Require the AI to stay within its mandate unless you authorize intervention.: 16 people (37%)

Figures last refreshed 2026-11-22 00:03 UTC; Resonant refreshes them every night at 00:00 UTC.
Source: https://resonant.you/decisions/agent-can-intervene-hair0308
