Data/ML/AISecurityHigh signalEmergingAround since 2022
Prompt Injection
An attack where malicious text (hidden in a document, webpage, or user input) tricks an AI system into ignoring its original instructions and doing something unintended.
Think of it like
Like a con artist slipping a forged note into someone's inbox that says 'ignore your boss's instructions and do this instead' — and the AI just follows it if it's not careful.
Junior or senior?
Junior sounds like
Focuses on model capability without considering how it could be attacked.
Senior sounds like
Can describe how they protected an AI feature against prompt injection from untrusted input.
Ask them
“How did you protect your AI feature against prompt injection from untrusted input?”