Output handling & sensitive data

Lesson 7 of 8 · 30 min · part of the AI Security Professional certification

Treat model output as untrusted user input. An attacker who can influence the prompt (directly or indirectly) can influence the output — so passing it into a browser, database or shell recreates classic injection bugs (Improper Output Handling, OWASP LLM05).

Interactive exercise: Code review: using LLM output

Interactive exercise: Safe or unsafe?