Item — HN Reader

Hopefully this is on-topic: I would hope that some people would opt for private chat services. I evaluated both Proton’s Lumo+ and DuckDuckGo’s Duck.ai services. I like both services but only wanted to pay for one and I chose Lumo+ because chat history is stored with my Proton data and is available in all my devices, Duck.ai stores chat history on current access device. Both services are also very usable with their free plans.

At least some of us in HN talk about limiting the data we give to Facebook, Google, Microsoft, etc. Isn’t it just as important to limit what we share with non-privacy preserving AIs?

Note: tech friends have asked me how I can use slightly weaker AI models and be happy about it: I still use Gemini Plus (and Anthropic via AntiGravity) for technical work: everything I do as a software developer is open source and all of my writing (20+ books) is Open Content so I don’t care about privacy and being direct-marketed based on my tech work. To me it makes sense to use the best AI just for tech work and a private AI for everything else. Think about this if a family member has a serious health problem, or something else private: do you want to use open web searches and open AI chats, or do you want to use private web search and private AI access? Why not make privacy your default, except in special situations?

3 comments

> DuckDuckGo's Duck.ai services ... what we share with non-privacy preserving AIs.

Duck.ai's Privacy Policy goes:

  As noted above, we call model providers on your behalf so your personal information (for example, IP address) is not exposed to them. In addition, we have agreements in place with all model providers that further limit how they can use data from these anonymous requests, including not using Prompts and Outputs to develop or improve their models, as well as deleting all information received once it is no longer necessary to provide Outputs (at most within 30 days, with limited exceptions for safety and legal compliance).

This is not much different to the BigLabs, tbh.

Otoh, privatemode.ai, confer.to, trymaple.ai are at least attempting Apple AI-like confidentiality.

That's a fair point, but DuckDuckGo has been a privacy champion for years, so I would give them far more weight in actually adhering to these policies as a middle-man than to directly trust the others. The priorities are different.

I have a family member with a serious health problem, and I've been using Claude code to put together a very comprehensive medical dossier.

I'm not worried about the privacy aspect though many suggest that I should be. The power the dossier has given them to navigate the medical industry in the United States has been absolutely incredible. They don't have to be stuck when a random doctor who has never heard of their illness suggests that they might be overreacting. They can simply find someone who will help them. They can talk, in medical lingo, about their test results and discuss them with the doctor on equal footing.

I'm not sure this would've been nearly as successful without Opus 4.5/4.6 driving the harness. I'm not also not sure what real privacy risk there is here; it all sounds very theoretical.

There are also encrypted AI chatbots like Tinfoil and Confer that E2E encrypt all data to a secure hardware enclave. I also use Claude and OpenAI for non-privacy needing tasks, but use Kimi-k2 on tinfoil when I need privacy. Kimi-k2 feels close enough to SOTA so I'm happy with it.