Manage: What Happens When AI Goes Wrong

Every fire drill feels silly until the day it doesn't. The Manage function is the fire drill of the NIST framework: deciding in advance what happens when your AI causes harm, makes a costly mistake, or just starts behaving strangely.

A real manage plan answers four questions. Who gets the call when something goes wrong? What do they do in the first 24 hours? At what point do we pause or shut the tool off? And how do we tell the people who were affected?

Notice what's not on the list: hoping it won't happen. Hope is not a control. The organizations that survive AI incidents aren't the ones with perfect models. They're the ones that practiced the response while everything was calm — and wrote it down, so the plan survives staff turnover.

I explain frameworks; I don't give legal counsel.