# Guardrails

> Rules, checks, filters, or workflows that keep AI behavior within allowed boundaries.

- Category: Evaluation & Safety
- Stage: Established
- Also called: Safety checks, Policy gates
- Canonical: https://www.joelcsaji.com/ai-dictionary/guardrails

## Plain English

Guardrails are the bumpers around what an AI can say or do.

## Example

A tool guardrail blocks an agent from sending an email to more than 20 recipients without approval.

## Why it matters

Guardrails are mandatory once AI systems can take actions, access data, or influence decisions.

## Related

- [Approval Gate](https://www.joelcsaji.com/ai-dictionary/approval-gate)
- [Action-Time Authorization](https://www.joelcsaji.com/ai-dictionary/action-time-authorization)
- [Prompt Injection](https://www.joelcsaji.com/ai-dictionary/prompt-injection)

## Mentioned in

- [Open Claw: The Rise, The Chaos, and What Comes Next](https://www.joelcsaji.com/blog/open-claw-rise-chaos-what-comes-next)

## Sources

- [OpenAI: Function Calling](https://developers.openai.com/api/docs/guides/function-calling)
- [OWASP Agentic AI Top 10](https://genai.owasp.org/2025/12/09/owasp-top-10-for-agentic-applications-the-benchmark-for-agentic-security-in-the-age-of-autonomous-ai/)