Skip to content
ML Academy

RAG and agents

Interview question

  • seniorlevel
  • 15 minsuggested time
  • 2Applied AI, ML engineer

You are building an email assistant agent that reads incoming customer emails, looks up bookings and can send replies and issue travel vouchers. How could prompt injection hurt you, and what controls would you put in place?

Tests: Threat modelling for LLM agents: direct and indirect prompt injection, data exfiltration, tool misuse, and controls that are enforced outside the model.

day 47day 49day 53

Sign in to practise: answers are saved before model answers are shown.