Can We Use DeepSeek Safely?

We have a new AI kid on the block, and he’s getting all the attention. DeepSeek competes with or beats OpenAI, claims to have done it on the cheap, with far fewer resources, and is out there as an open-source option. It’s one of the top downloaded apps, and it forced the US tech market to take a hit as investors questioned why they spent billions on AI models when China could do it at a fraction of the cost. 

And there’s the problem...

How can we trust an LLM from China with no caution? The majority of LLMs offer a “privacy” option that allows you to prompt without sending all your data to be trained on or stolen and used as they see fit. DeepSeek currently offers no such functionality. On the contrary, it states that it collects personal information, chat histories, prompts, audio files, and uploaded documents and does not clarify how that data will be processed or how sensitive information will be safeguarded. 

So here we are, AI becoming ubiquitous, and we have a top-tier LLM that’s able to run well on lesser hardware with less energy but is also going to steal all your content, your prompts, and any of the documents you load into it. Oh, and the Chinese government will also do whatever it wants to with your data. I know we’re no longer reading the fine print, but this one is pretty straightforward. We should be a little cautious.

What are we going to do? We’re going to use it, that’s what. But rather than exposing your data, take advantage of the benefits of this and any future models like DeepSeek safely, securely, and with no risk of data loss inside Backplain. 

Backplain is adding DeepSeek to its already expansive AI model list in a unique way. Your favorite AI model aggregator is putting your privacy and data first by containing the DeepSeek-R1 LLM. The only input into DeepSeek through Backplain will be Backplain’s own API and the only egress will be to the Backplain output box. Any attempt by the model to communicate outside of the Backplain infrastructure will be stymied, blocked, or denied. 

Ultimately, Backplain will do what it was designed to do, prevent your data from being trained on, used by, or delivered to any entity. We are offering DeepSeek safely on Backplain University for all subscribers. For our Backplain Enterprise companies, we are including the DeepSeek LLM as an optional LLM should you decide to deploy it.

Backplain allows you use LLMs like DeepSeek without the worry of data loss. We want you to have all the benefits of AI while protecting your data.