I Built a Failure-Mode Test For OpenAI Privacy Filter. The Model Card Was Right.

Towards AI
Generative AI Open Source AI AI Research

A small enterprise-style test showing why PII redaction needs domain rules, span cleanup, and in-domain evaluation before it becomes a production control. Most sensitive values get caught. The gap is the production problem. Image: AI-generated by the author. At work, this problem comes up in a very practical way. Teams often need to send large user transcripts, messages, or operational notes to third-party APIs.