Skip to main content
AI & Agents

AI Guardrails

Guardrails are the constraints that limit what an AI system is permitted to do, enforced outside the model rather than requested inside a prompt. They include permission-scoped tools, spend and rate limits, blocked destructive operations, output validation, and approval gates on irreversible actions.

The defining property of a guardrail is that it holds when the model is wrong. An instruction in a system prompt asking a model not to exceed a refund limit is a preference; a server-side check that rejects the call is a control. This distinction is the single most common gap between agent demos and agent deployments, and it is invisible until something goes wrong.

Guardrails also need to survive manipulation. Content returned by a tool — a support ticket, a web page, a document — can contain text aimed at the model, so tool output should be treated as untrusted data rather than instructions. An agent that is talked into calling a destructive tool still cannot call one that was never exposed to it.

FAQ

AI Guardrails
FAQ.

Common questions about ai guardrails.

Ask Us Anything

Not reliably. Prompt instructions shape behaviour but do not constrain it, and they can be overridden by injected content or simply ignored under unusual conditions. Real guardrails are enforced in the tool layer and checked server-side on every call, so they hold regardless of what the model decides. Prompt guidance is useful in addition, never instead.