Paying 10 dollars a month to have a private wrapper for mainstream LLMs.
Chats are retained anyways, it just won’t be tied to your real identity. I don’t see the point of doing this unless you trust DuckDuckGo with knowing it instead of OpenAI or Anthropic?
Not much different in some sense than trusting your VPN provider.
This seems to be an emerging (?) issue in the privacy space, there is a lot of skepticism towards AI services (see Lumo thread) yet they’re used by a lot of people. There is an option to run them locally on your machine, in practice this is quite difficult to do on the higher parameter models unless your running a high end machine. I also like NanoGPT for this use case.
As with anything, considerations must be made in terms of your threat model.
It you want real privacy, run the LLM locally with Ollama or similar tools.