gensai.aiGuardrails are the mechanisms that bound an AI system's behaviour: input and output filtering, limits on permissions and actions, human validation of sensitive operations, refusal rules. They turn a probabilistic model into a system whose limits can be defined.
In practice at Gensai
Every Gensai agent ships with its guardrails: minimal access rights, an explicit list of allowed actions, human validation on critical steps.