Kinda crazy how many people submit their sensitive information to frontier models. Tax documents, health data, financial transactions etc.
Built this as a proof of concept. Venice AI is on a similar wavelength where they limit identity connected to prompts, but not information in the prompts.
Sometimes it's tough to actually filter all sensitive information and keep the same quality in responses. I can share the github if anyone is interested.
Note: Was also experimenting with OpenMed Privacy Filter on my mac, but it is not compatible with a linux server.
Comments URL: [https://news.ycombinator.com/item?id=50160374](https://news.ycombinator.com/item?id=50160374)
Points: 1
source & further reading
privai-sandy.vercel.app — original article