Incident Response Plan for an AI Mistake
Prepare steps for when an automated output causes harm or embarrassment.
Added October 15, 2026. Written with AI assistance and checked for clarity and placeholders; not reviewed by a lawyer, accountant, clinician or other professional. How these prompts are made
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
[VERIFY] is an editorial flag, not a blank to fill. Wherever the prompt shows [VERIFY], check that item against a reliable source before you use it (what [VERIFY] means).
Write an incident plan for mistakes made by our AI-assisted process: [PROCESS_DESCRIPTION]. Possible harms: [POSSIBLE_HARMS]. People to inform: [STAKEHOLDERS]. Tools to pause: [SWITCHES]. Create phases: detect, contain (how to pause the automation), assess (who was affected, what was sent), correct (message to those affected), learn (cause and fix). Provide roles, a decision tree for severity levels, and draft wording for a short apology and correction under 100 words. Add a log template and a rehearsal scenario with invented details to practise in 30 minutes. Mark notification duties as [VERIFY] for [JURISDICTION]. Also include a short communication rule for staff: who may speak publicly, what to say to a customer who calls about the incident, and a requirement to preserve inputs and outputs before anyone edits or deletes them. Close with a checklist for the first hour after discovery, with five actions in order, a named role for each, and a place to record the time each action was completed for the later review.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.