Guide
Glossary
Guardrails
Guardrails are the rules and technical controls that limit what an AI system can see, say and do, so it stays within what an organisation has approved. They include restricting which documents and systems the AI can reach, blocking certain topics or data such as tax file numbers, checking outputs before they are used, limiting what an AI agent may do without a person's approval, and recording every action in an audit log. Some guardrails are settings in the AI product; others are built into the workflow around it. No single guardrail is foolproof, so they are layered. Good guardrails are specific and testable, such as "no email leaves the firm until a named person approves it".
Also called AI guardrails, AI safety controls
Last updated:
Example
In a law firm
For example, a family law firm uses an AI agent to draft replies to new enquiries. Its guardrails: the agent can read the enquiry inbox but not matter files, it must not give legal advice, every draft waits for a lawyer's approval before it is sent, and each action is logged so the practice manager can see what the agent did and when.
Guides that explain it in context
Secure by design. Set up correctly. Fully managed.
Talk to us before you commit to anything
Start with a free 45-minute discovery call. We look at your systems and priorities, then recommend a first step with a fixed scope, or tell you if we are not the right fit.