"We do not save or store your Prompts or Outputs.
Additionally, all metadata that contains personal information (for example, your IP address) is obfuscated from underlying model providers (for example, OpenAI, Anthropic).
If you submit personal information in your Prompts, it may be reproduced in the Outputs, but no one can tell whether it was you personally submitting the Prompts or someone else.
We have agreements with model providers to further protect your privacy.
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 that includes not using Prompts and Outputs to develop or improve their models as well as deleting all information received within 30 days."
> so your personal information (for example, IP address) is not exposed to them
Is this not two different things? The first one saying that the IP address is changed in some way to make it unidentifiable. The second one is saying it's just not sent.
Metadata that contains PII is not shared.
No mention of just data with PII, i.e. your chats.
You aren’t providing communications content privacy, you’re providing a meta data proxy which is not remotely the same.
If you submit personal information in your Prompts, it may be reproduced in the Outputs, but no one can tell whether it was you personally submitting the Prompts or someone else.
What else should they call it? I suppose "anonymous" could work (i.e. they only know about you what you tell them).
Either use one of the many open models that beat GPT-3.5 and you have a real, and better!, product.
I literally have no idea what the point of this is, or why not make the actual step of rolling out a real private LLM. It’s literally inferior to what Mixtral provides right now, under Apache license.
In their golden era search engines were essentially super sophisticated keyword matching. Users used to find novel websites from passionate people. Now it's becoming more of a social media feed of the same 100 website where you train it what you like.
This is exactly why search is so vulnerable to disruption from LLMs.
Citation? I can't find this.
>to hiding websites in search due to political bias
This is a gross mischaracterization. Every search engine down weights domains and pages with bad or inaccurate information. Their entire purpose is to find good information, not penis enlargement 4 u .ru pages.
For "hiding websites in search", they may be referring to DDG downranking russian propaganda sites during the russian-ukraine war: https://www.pcmag.com/news/duckduckgo-to-down-rank-sites-ass...
never saved by us, and not used to train AI models
could make the more paranoid believe that they could be saved by DDG’s partners - from very directly (Anthropic saves chats with Claude) to indirectly (all channeled through a shady arms-length subsidiary).