how do you use frontier models without exposing your data, such as personal records, company files, financial documents, or confidential information?
Commit
7138440
I recently needed to work with a lot of customer data to improve performance at large scale. I dictated the llm to write a script that rewrites all personal data in the db with random names/strings/whatever. Worked out great! But probably not worth it for short tasks
a local LLM? An Italian developer has built an interesting project for Italian legal text: https://github.com/Rizzo-AI-Academy/rizzo-pii (a small CPU-only model that anonymizes documents locally before you send them to an LLM)
GitHubGitHub - Rizzo-AI-Academy/rizzo-pii: Local-first privacy guard: anonymize your documents before sharing with LLMs.Local-first privacy guard: anonymize your documents before sharing with LLMs. - Rizzo-AI-Academy/rizzo-pii
commit 7138440