Apply incident-response sequencing to an AI system that may have leaked sensitive data or taken unsafe action.
Data leak A support AI returns another tenant's account summary. The queue is live, the customer has screenshots, and the team does not yet know whether this is a retrieval bug, permission bug, or prompt leak. The first minutes decide whether the team keeps enough evidence to scope the incident. NIST incident response applied to AI Contain, preserve, scope, eradicate, recover. AI changes the artifacts, not the discipline. You still reduce harm first and keep enough evidence to understand impact. Delete and reset Looks clean, destroys proof Lower exposure with a defensible evidence trail. Do not clean up the scene…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in