Guardrail

A guardrail is a preventive, detective, or corrective control intended to keep system behavior within defined boundaries.

Plan for this page

Understand one concept in plain language and connect it to nearby ideas.

You will leave with: Guardrails include input validation, access limits, content filters, policy checks, output validation, approvals, rate limits, monitoring, and recovery. Each addresses particular failure modes and can create false positives, false negatives, or new dependencies.
Time: 3–8 min
Before you begin: No prerequisite. Keep one real example in mind.
Do this now: Read the explanation, then compare the example with the common misunderstanding.

Prepared next stepBuild layered agent controls

GuardrailA guardrail is a preventive, detective, or corrective control intended to keep system behavior within defined boundaries.

Alignment
Recovery
Security

Plain explanation

Guardrails include input validation, access limits, content filters, policy checks, output validation, approvals, rate limits, monitoring, and recovery. Each addresses particular failure modes and can create false positives, false negatives, or new dependencies.

Example

A publishing agent cannot post externally until a policy checker passes and a named editor approves the final destination and content.

Why it matters

No single guardrail makes a system safe. Layered controls should match actual hazards, be tested against bypass and failure, and include a response when they trigger.

Common misunderstanding

Misconception: guardrail is another word for prompt. Correction: prompts can be one layer, but enforceable controls also live in software, permissions, process, and monitoring.

Related concepts

Follow relationships, not only the alphabet

AlignmentAlignment is the effort to make an AI system’s behavior accord with specified human intentions, values, rules, or objectives.RecoveryRecovery is the ability to stop, contain, reverse, retry, compensate for, or safely continue after failure or interruption.SecuritySecurity protects systems and information against unauthorized access, change, disclosure, disruption, or destruction.

Build layered agent controls

Sources and scope

What supports this object

Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence ProfileU.S. National Institute of Standards and Technology

Provides voluntary actions for testing and managing generative-AI risks; individual safeguards remain fallible and context-dependent.

Open source

Loading page content...

…