n8n Guardrails Redacts PII Permanently, Leaving Gaps in Real Workflows

n8n's Guardrails node, launched in November, masks personally identifiable information before it reaches an LLM, but permanently deletes the original values rather than storing them. This creates a practical problem for workflows like support automation, where the real PII — such as a customer's name or email — is needed to write meaningful records to a CRM. The design is intentional: Guardrails is built as a content moderation tool, and discarding sensitive data is the correct behavior for jailbreak or NSFW filtering. However, that same behavior makes it unsuitable for workflows requiring PII to be recovered at a later, trusted stage. A tool called Privent is proposed as an alternative, using reversible tokenization so that masked values can be restored at a designated output point without ever exposing them to the LLM.