DeepSeek AI is the newest app from China to go viral. It currently sits at the top of the App Store, where it displaced its direct rival ChatGPT. People interested in AI apps are flocking to try out this AI alternative, likely after hearing that the Chinese developer managed to create a reasoning DeepSeek AI as good as ChatGPT o1 without having access to the same massive infrastructure as US AI firms.
DeepSeek had to rely on a limited number of high-end NVIDIA GPUs to train the AI. Most of them were probably smuggled into the country, considering that Chinese startups are banned from acquiring key chips needed to power AI. DeepSeek’s AI innovations tanked the market on Monday, as some investors worried that competitive models like R1 don’t necessarily need high-end NVIDIA GPUs and similar chips.
If you’re excited about the prospect of trying something different than ChatGPT, Gemini, Claude, and Meta AI, you’ll probably want to test DeepSeek as soon as possible.
You’re most likely to run the iPhone or Android DeepSeek AI apps, or go through the web interface on a computer. If that’s how you’re going to access DeepSeek, you should know that all your data, including your chats, will be sent to China and stored on Chinese servers.
If TikTok has taught us anything, it’s that having treasure troves of user data stored in China for possible exploitation will trigger regulatory action. That’s not to say the US government will ban DeepSeek anytime soon. But it goes without saying that you should pay attention to which AI services you use and how your data is handled.
The best-case scenario for good DeepSeek privacy
I’ve been using ChatGPT for two years, and privacy has always been my main concern. I want to know what happens with my data, and I don’t want any of my chats to train the AI. OpenAI now offers opt-out options, but those weren’t there from the get-go.
With DeepSeek, the problem isn’t just ensuring user privacy is respected. It’s trusting that the user data is stored safely in China and that the local government will not want access to it.
Before we look at all the types of data that DeepSeek collects, you should know there is a more private way to run the DeepSeek AI models. You’ll have to install them and run them locally on your laptop. In this case, DeepSeek won’t collect user data from you because you won’t be talking with its servers when using the AI.
That’s not necessarily something most people will want to do. And you can’t run DeepSeek V3 and R1 on your iPhone or Android phone.
The most obvious privacy problem
If you’re going to use a mobile phone or the web interface, this is the most important thing to know about DeepSeek privacy: it stores all your data in China. The company says as much on its privacy page:
The personal information we collect from you may be stored on a server located outside of the country where you live. We store the information we collect in secure servers located in the People’s Republic of China.
Also, the data is stored for “as long as necessary” to provide DeepSeek services and serve other purposes.
What user data does DeepSeek collect?
To use DeepSeek, you’ll need to create an account. Once that happens, DeepSeek will start collecting information about you as follows:
Profile information. We collect information that you provide when you set up an account, such as your date of birth (where applicable), username, email address and/or telephone number, and password.
User Input. When you use our Services, we may collect your text or audio input, prompt, uploaded files, feedback, chat history, or other content that you provide to our model and Services.
Information When You Contact Us. When you contact us, we collect the information you send us, such as proof of identity or age, feedback or inquiries about your use of the Service or information about possible violations of our Terms of Service (our “Terms”) or other policies.
The company also collects information automatically, including information about how you access DeepSeek and what devices you use. It even collects “keystroke patterns or rhythms.”
If you access DeepSeek from multiple devices, it’ll associate them under your account. There’s also usage information, cookies, and payment information.
Then there’s user data coming from “advertising, measurement, and other partners:
Advertisers, measurement, and other partners share information with us about you and the actions you have taken outside of the Service, such as your activities on other websites and apps or in stores, including the products or services you purchased, online or in person. These partners also share information with us, such as mobile identifiers for advertising, hashed email addresses and phone numbers, and cookie identifiers, which we use to help match you and your actions outside of the Service.
All of that sounds similar to the privacy policy of any app, whether it’s from China or other countries. But when it comes to AI apps, privacy policies also cover the data collected from interactions with the chatbot. This is where users might provide all sorts of information to the chatbot while talking to it. I’ve always reminded you to avoid sharing sensitive information with any AI product, and the same goes for DeepSeek.
Yes, the privacy policy mentions that DeepSeek will collect prompt data:
Review, improve, and develop the Service, including by monitoring interactions and usage across your devices, analyzing how people are using it, and training and improving our technology.
The privacy policy does mention user rights, including “a right to access, change, oppose, request a copy of your authorization, file complaints before the competent authorities, withdraw your consent, or limit our collection and use of your personal information as well as to request that we delete it, and potentially others.”
You can also opt out of training. You can delete your chat history via the app’s settings.
As for the partners that DeepSeek might share user data with, Wired found that DeepSeek sends data to Baidu and Volces.
The DeepSeek privacy policy also says it’ll use data to “comply with legal obligations,” which is typical for such documents.
The big elephant in the room
While the privacy policy seems on par with what you’d expect from OpenAI, Google, Anthropic, and Meta, there is a clear distinction here that I’ll underline again. All data is sent to China, and transparency isn’t perfect. We have no way of knowing whether DeepSeek is indeed protecting user data as best it can or whether it’ll give the Chinese document blanket access to said data.
It’s TikTok all over again, but at a larger, potentially more dangerous scale. Remember that DeepSeek AI isn’t perfectly transparent even if it’s pacakged as open-source. Real-world examples have shown that DeepSeek will refuse to answer questions about certain events in China’s past or present. Censorship has been used while training the AI.
Also, the TikTok divest-or-ban law wasn’t just about the location of user data. US officials were also worried about the TikTok algorithm that can influence users on all sorts of matters, including their stance on China.
Who is to say that DeepSeek doesn’t include instructions the end-user can’t see to potentially manipulate the user beyond the expected hallucinations an AI chatbot might deliver? Even if DeepSeek doesn’t include a nefarious algorithm, DeepSeek is likely just one of the many competitive AI products coming from China that will be available in the US and other Western markets.
With all that in mind, you should be very mindful of what data you trust DeepSeek with. Meanwhile, I’m fairly certain that regulators in the US, EU, and elsewhere are already looking at these developments with plenty of concern.
The post DeepSeek AI collects tons of data about you and sends it all to China appeared first on BGR.
