What are guardrails, and which safeguards should run in code?
Review note: The answer refers to the original author's CV or target job offer — adapt it to your own experience.
Answer out loud or in your head first, then compare.
Answer
Guardrails are limits and checks meant to keep the system working within agreed boundaries. Some can rely on a model's judgement, but things like permissions, allowed addresses, maximum amounts and valid arguments should be enforced by code.
Example: the model proposes a refund. The backend checks the user's right and the refund limit itself. It doesn't grant them just because the model wrote “approved”. Safeguards need testing too, because over-blocking gets in the way of work. Another model can help detect fuzzy risks, but it also makes mistakes. What matters is limiting what can be done even when such a judgement fails.
Why it matters: Explains an important term from the offer through concrete responsibility.
Possible follow-ups: What blocks an action before it runs, and what only detects a problem later?
How did it go? Be honest — “Not yet” puts the question on your review list.
Start with
- H-P01 · How does an AI agent differ from a plain model call? — Agents and workflows
- H-P02 · What is a workflow, i.e. a fixed sequence of work? — Agents and workflows
Further reading
- OWASP: LLM Prompt Injection Prevention Cheat Sheet(opens in a new tab)Docs; accessed 22.10.2026. Threat model, untrusted content, tool control and defence in depth.