Open cardosec

Case ABCD5E · AI security · L2 Practitioner

Improper Output Handling

Practise as: Explain it
  1. 01 What is it?
  2. 02 How is it abused?
  3. 03 How do you stop it?
Practise this card
What a strong answer covers

Try it out loud first. Then check yourself:

  1. OWASP LLM05: LLM output passed to downstream systems without validation, like any untrusted user input.
  2. Rendered as HTML/markdown it can cause XSS; passed to a shell, eval or SQL it can cause RCE or injection.
  3. Generated URLs fetched server-side can cause SSRF, e.g. to the cloud metadata endpoint 169.254.169.254.
  4. Fix: context-aware output encoding, parameterized queries, strict JSON schema validation, sandboxed execution.
  5. Combines with prompt injection: attacker controls input, model echoes payload, app executes it.

If the interviewer pushes back

  • How would you safely let an LLM generate and run SQL for analytics users?

cardosec draws a security topic and gives you a clock: explain it out loud with no notes, then see what you covered and what you missed. Free during early access.