Agents & Agentic AI
Guardrail
A technical or policy control that limits what an AI system can do.
Example
A guardrail blocks an assistant from sending messages containing unapproved confidential data.
Why it matters
It helps limit unwanted behavior while keeping useful work possible.
You'll hear it
“That rule stopped the risky action but also blocked a normal request.”
Can you control it?
Sometimes
Sometimes. Some products or workflows let you influence a guardrail, while others keep it fixed or manage it automatically. Check the settings and documentation for the tool you are using.
Common questions
- Can a guardrail block something harmless?
- Yes. Broad checks can reject legitimate requests as well as harmful ones.
- Does passing one check prove an action is safe?
- No. A check covers particular risks and can miss others.
- Can guardrails become outdated?
- Yes. New tools, uses or attack methods can expose gaps in earlier protections.