Guardrails
Guardrails are the rules, filters and checks around an AI system that block harmful, off-topic or unsafe outputs and actions. They can sit inside the model's training, in its system prompt, or in separate software that screens inputs and outputs.
In one line, for a 12-year-old
Guardrails are the safety rails that keep an AI from saying or doing harmful things.
An example
Our Young Lab has no open AI chat at all — the strongest guardrail for children is not offering the risk in the first place.
Why it matters to people
Good guardrails protect people without blocking legitimate questions. Ask any organisation using AI with the public what its guardrails are and how they were tested.