Sort each AI failure mode into the control bucket that would most directly reduce it. Accuracy control Privacy control Security control Fairness control Accountability control The model invents a customer quote that is not in the source notes A prompt includes employee names and health details that the model does not need A user prompt tells the bot to ignore system instructions and reveal hidden context A lead score consistently deprioritizes one region without a business explanation No one knows who approves the AI-generated customer message before it is sent The answer cites a report title that does not exist…
Read the full lesson
Sign up free: one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in