Prompt Injection and Misuse

Talk it through with a coach who asks guiding questions and gives hints until you reach the answer. Pass to mark the challenge as completed.

The problem

  • hard
  • Pro

Your AI assistant can read a user's emails and take actions like sending replies. A security researcher shows that a malicious email can trick it into forwarding private messages. What do you do?

Talk this problem through with an AI coach

This is a Pro challenge. Sign in with Google, then upgrade to start a voice session.

Start free →