CPII: scrubbing personal data and honouring the right to be forgotten
A cache of LLM traffic is a cache of whatever users typed — including PII. CPII reports what personal data is present and executes right-to-erasure, so compliance is a command, not a project.
Any cache sitting in front of an LLM accumulates whatever users typed, and users type personal data — names, emails, account numbers, the contents of support tickets. That makes the cache a system of record for PII whether you intended it to be one or not, with all the obligations that follow. CPII is the tooling that makes those obligations executable.
CPII REPORT surfaces what personal data is present, turning 'do we have PII in the cache?' from an audit into a query. CPII ERASE executes right-to-erasure — when a user invokes their right to be forgotten, you can actually carry it out across the cache instead of hoping TTLs eventually cover it. Erasure that you can prove is the difference between a compliance posture and a compliance hope.
Five stages score every write before it can ever be served.
It composes with the privacy defaults that ship turned on: query previews are kept out of logs unless you opt in, structured logs can omit prompts, and retention windows bound how long anything lingers. CPII is the active layer on top of those passive defaults — not just 'we don't log much,' but 'we can find and remove exactly this person's data on request.'
The bottom line
Privacy regulation treats a cache like any other store of personal data, and rightly so. CPII gives a team the two verbs that regime actually requires — show me what's here, and remove this person — as commands, so compliance is something you run, not something you schedule a quarter for.