Constraints applied at the input, output, or system level to prevent unsafe, off-policy or off-brand AI behaviour. Includes content filters, retrieval guardrails and structured-output schemas.
This definition is maintained by Moweb partners and used in live client engagements. For how Guardrails applies to your estate, or to challenge a working definition, speak to a partner.