“Apprentice's platform includes optional 'guardrails' that require a human operator to approve or reject actions proposed by the AI agent before they are executed.”