Hi, interested in understanding what people’s main concerns are for their privacy when using AI.
The main issues as always are sending data off to a server somewhere. There’s been a big push toward making AI run locally which is great, and for requests to servers there’s been a lot of great work toward privacy on that front like homomorphic encryption, OHTTP, multi-party computation, and lots of other privacy enhancing technologies (PETs). A lot of work is being done right now on making server-side processing more private and secure, you can read up on it at the Confidential Computing Consortium. It’s really very exciting stuff, I think we’re seeing the dawn of a new era of privacy online. I guess the main issue is it’s not all super widespread yet so you can’t be completely confident that whatever AI you’re using is doing any of that, but you can always stick to local AI models if you want maximum privacy.
2 Likes