Show HN: Make your AI conversations private. ProtonMail equivalent of ChatGPT
I'm Ewan, a software engineer based in Switzerland and I've just finished the beta of Cognos, my attempt to bring privacy to your generative AI chats.
ChatGPT and co. (Gemini/Poe/You/Anthropic etc.) are powerful tools but they offer little privacy to you. Employees can read your chat history; your messages can be leaked (happened already to OpenAI); your data will be used to train their models and of course, hackers may get access to everything in the future.
Cognos sits in the middle and encrypts your messages and the AI generated responses. It's important to note this is not end-to-end encryption like Signal/WhatsApp as AI models need access to your plaintext messages. Instead we work like ProtonMail does for email and encrypt your message and the AI generated response as soon as we can. After this step, nobody but you can access your messages.
I'm sure - in true HackerNews style - there are some of you thinking "why would anyone need this? You can have more privacy with 2x3090s, Open WebUI + Syncthing", and you would be right. For the truly paranoid running everything yourself is likely the best option. But for those that don't have the $1k + running costs, or the knowledge to run it, or the desire to do so; I would like you to consider Cognos as a privacy friendly alternative.
Right now the beta is live (and free but rate limited) with some commercial and open-source models to choose from. The launch blog article linked below[0] gives more technical detail on the security side of things as well as some screenshots about what you can expect. If you want to jump straight in and sign up you can go to https://app.cognos.io/
Feedback is greatly appreciated so let me know if this is valuable to you, my email is below and my Threema ID is in the article.
Many thanks and happy hacking, Ewan
ewan@cognos.io