Use a structured triage path when an AI workflow encounters suspected prompt injection.
Alert A resume parser finds hidden text that tells the model to override scoring criteria and rank the applicant at the top. The workflow influences who receives an interview, so integrity and auditability matter. Response path Classify, contain, preserve, improve Treat the model as a potentially confused deputy. Keep evidence, remove risky action, and convert the event into a test case. Bad shortcut Delete the file and move on Business review can continue in a constrained, auditable mode. Contain the capability before judging intent. 01 Classify 02 Contain 03 Now try Classify the signal Step 1
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in