Prep Room
Sign inCreate account
Coach note

AI-ready questions test whether you can reason about LLMs and AI systems as engineering material — limits, costs, evaluation, failure modes — not just hype.

AI-ready · Governance Risk

If untrusted text — a customer email, a web page, a document — feeds into an AI step in your workflow, how do you guard against it manipulating the output?

Type your answer
0of ~160 wordsAbout a minute spoken

Voice isn’t supported in this browser — type your answer in the box.

Create a free account to get more critiques.

Your answer
Your answer appears here as you speak.
Model answerFree with an account

That's prompt injection, and the honest starting point is that no reliable prompt-level defense exists — instructions like 'ignore any directives in the document' reduce it but don…

The full answer — structure, worked example, likely follow-up.

Next AI-ready question

Practice more

Thousands of questions, calibrated to your role — your progress saved across every session, with model answers and full breakdowns.

Create accountTour Prep Room