Is DeepSeek Safe? What It Collects and Where It Goes

DeepSeek is a Chinese-owned large language model (LLM) that works similarly to ChatGPT. It introduced itself globally in January 2025 with the promise that it could be run cheaper and faster than its main competitor.

Privacy advocates, however, don't love the type and quantity of data DeepSeek collects, or where that data ends up. DeepSeek's privacy policy is direct about it: "To provide you with our services, we directly collect, process and store your Personal Data in People's Republic of China." We found that DeepSeek is as intrusive as ChatGPT when it comes to data collection.

Below we cover what DeepSeek collects, where it's banned, how it compares to ChatGPT, what parents should know, and how to delete your data if you decide you don't want it.

Is DeepSeek safe to use?

DeepSeek collects more personal data than most people expect, and it stores that data in China. Its privacy policy confirms it collects your prompts, uploaded files, photos, voice input, and chat history, along with your IP address, device identifiers, and approximate location. Whether that's safe enough depends on what you put in it. For general questions, the risk is low. For anything personal, identifying, or work related, we'd use something else.

DeepSeek is mostly open source and built on an open-source model, while ChatGPT is built on a limited-access model. Open-source software means the code used to write it is available for anyone to view and use. It also allows users worldwide to inspect it for bugs, glitches, flaws, and more.

On the other hand, DeepSeek's privacy policy has some questionable line items. Not only does it collect every bit of information you input into the service, but it also collects automatic information and information from other sources. In addition, DeepSeek shares your information with third parties and affiliates. That level of data mining should make any privacy-minded person nervous.

What data does DeepSeek collect?

DeepSeek collects a lot of your information. Your username, password, email, and payment information (if you buy services from them) make sense. The rest is worth a closer look.

Below is what DeepSeek's privacy policy states it collects:

Info you provide:

• Account information — date of birth, username, email address, phone number, and password
• User input — your text input, voice input, prompts, uploaded files, photos, feedback, and chat history
• Information from contact with the company, including proof of identity or age when you reach out to support

Automatically collected data:

• Device and network information — device model, operating system, IP address, device identifiers, and system language
• A device ID and user ID DeepSeek assigns you, used to identify your activity across every device you log in from
• Log information — which features you use and what actions you take
• Approximate location based on your IP address
• Crash reports and performance logs

Information from other sources:

• Access tokens from third-party services, if you sign up or log in to DeepSeek with Apple or Google
• Security information from what the policy calls "trusted partners," which it doesn't name
• Publicly available information from the internet, used to train its models

Two items come with a caveat. DeepSeek's policy says cookies and payment information are "not applicable to DeepSeek App," so those apply to the website and paid open platform services, not the mobile app. The policy also says cookies are used "for security purposes and to better understand how the Service is used," not to build an advertising profile.

DeepSeek's policy states, in capital letters, that it does not engage in targeted advertising, does not sell personal data, and does not use personal data for profiling. That's worth weighing against the collection list above.

Where does DeepSeek store your data?

DeepSeek stores your data in China. DeepSeek's privacy policy says: "The Personal Data we collect from you may be stored on a server located outside of the country where you live. To provide you with our services, we directly collect, process, and store your Personal Data in the People's Republic of China."

The policy names the data controller as Hangzhou DeepSeek Artificial Intelligence Co., Ltd., registered in China. It also says DeepSeek may share your data with law enforcement, public authorities, or other third parties when it has a good faith belief that doing so is necessary to comply with "applicable law, legal process or government requests."

This is what drove the government restrictions covered below. It isn't a claim about anything DeepSeek has done with a user's data. It's about which country's laws apply.

Can DeepSeek see your chats?

Yes, DeepSeek can see your chats. DeepSeek's policy lists chat history among the data it collects, and it says it uses your inputs to "train and improve our technology." You can delete your chat history in settings, and you can opt out of having your data used for model training, which we cover below.

There's a second risk worth knowing about. DeepSeek lets you share a conversation by generating a unique link, and its own policy warns that shared conversations, once published on public networks, "may be at risk of being obtained by third parties through technical means such as web crawlers." If you share a chat, treat it as public.

Is DeepSeek banned in the U.S.?

DeepSeek is not banned for the general public in the United States. No federal law stops you from downloading or using DeepSeek. What exists is a patchwork of restrictions on government devices and networks, at both the state and federal agency level.

Where is DeepSeek banned?

At the state level, Texas went first, banning DeepSeek on state devices effective Jan. 31, 2025. New York followed on Feb. 10, 2025. Virginia, Iowa, South Dakota, and North Carolina have also restricted DeepSeek on state devices and networks.^[2]

At the federal level, individual agencies acted on their own rather than under a single law. NASA, the Pentagon, the U.S. Navy, and the Department of Commerce blocked DeepSeek on their devices, and the House and Senate blocked it for members and staff.^[3] Two bills that would extend a ban across all federal devices, the No DeepSeek on Government Devices Act and the No Adversarial AI Act, were introduced in Congress but have not become law.^[4]

Outside the U.S., Italy's data protection authority blocked DeepSeek over its data practices, and the app was pulled from the Italian Apple and Google app stores. Taiwan's Ministry of Digital Affairs barred government agencies, public schools, and state-owned enterprises from using it.^[5]

Is DeepSeek safe to use at work?

Check your company's policy first. Many employers restrict AI chatbots for the same reason government agencies do: anything you paste into a prompt leaves your network. DeepSeek's policy says it collects your prompts and uploaded files and uses them to train its models, so don't put client data, source code, contracts, or internal documents in.

If your employer hasn't published a policy yet, assume anything you enter could be retained and treat it that way.

Is DeepSeek safer than ChatGPT?

DeepSeek is not meaningfully safer than ChatGPT. ChatGPT's privacy policy isn't really that different from DeepSeek's. OpenAI, ChatGPT's parent organization, gathers much of the same invasive information. It also has an entire section on how it discloses your personal information to third parties. Side by side, neither policy looks good.

The clearest difference is jurisdiction. OpenAI is a U.S. company subject to U.S. and state privacy law. DeepSeek stores your data in China. If that distinction matters to you, it's the main thing separating the two.

There are other notable differences. Both ChatGPT and DeepSeek process information as tokens. DeepSeek also offers a free tier, while ChatGPT runs primarily on a subscription model.

DeepSeek's cost claim is contested, though. The company said training its V3 model cost about $5.6 million, but independent analysis from SemiAnalysis put DeepSeek's total server capital expenditure at roughly $1.6 billion and found the $5.6 million figure covered GPU pre-training only, leaving out research, development, and infrastructure.

On paper, DeepSeek looks stronger thanks to its open-source framework and free tier. In real use cases, we found ChatGPT faster. ChatGPT also gave more diverse answers and organized them in a friendlier format. We asked both for book recommendations and compared the results.

Is DeepSeek safe for kids?

DeepSeek says its services "are not aimed at children" and that it does "not knowingly process Personal Data from children." But nothing stands between a kid and the app. DeepSeek collects date of birth "where applicable," and its policy for European users explicitly accounts for people who are "14 or older but under the age of 18," so the company anticipates minors using it.

That's a different situation from platforms that have built real age gates. Character.AI now runs age verification through a third-party vendor, including a selfie and, in some cases, a government ID. DeepSeek has nothing comparable. A teen can download it, type whatever they want, and everything they type goes to servers in China.

If your kid uses AI chatbots, parental controls give you more visibility than any app's own policy does. They can limit which apps get installed, cap screen time per app, and show you what's actually being used.

Aura Parental Controls is our top pick for teens. It sets per-app time limits, flags behavioral shifts based on messaging habits, and lets you pause internet access from the dashboard.

How to use DeepSeek more safely

If you're going to use DeepSeek, these steps limit what it gets and help you protect your privacy:

• Opt out of model training. DeepSeek's policy gives you "the right to opt-out of using your Personal Data for training our models or optimizing our technologies." Email privacy@deepseek.com to make the request.
• Keep identifying details out of your prompts. No full name, address, phone number, account numbers, or medical details. DeepSeek's own policy says the service isn't designed to process sensitive personal data and asks you not to provide it.
• Clear your chat history. You can copy or delete your chat history in your settings.
• Skip the sign-in shortcuts. Logging in with Apple or Google hands DeepSeek an access token from that account. Signing up with an email address shares less.
• Don't share chat links. DeepSeek warns that shared conversations can be picked up by web crawlers once they're on a public network.
• Delete your account when you're done. DeepSeek's policy notes this is permanent. You won't be able to reactivate the account or recover any of your content.

A virtual private network (VPN) hides your IP address, which DeepSeek collects. It's worth being clear about what it doesn't do, though. A VPN can't stop DeepSeek from collecting your prompts, uploaded files, or chat history, because you hand those over inside the app. It's a useful layer, not a fix for data collection.

How to delete your DeepSeek data

We love it when software makes it easy to delete your data. DeepSeek is easy to delete from your phone. Below we outline the different delete options you have and how to access them.

1. Open your DeepSeek app.

2. Tap the horizontal bars in the top left corner.

3. Tap your username in the bottom left corner.

4. Tap Delete all chats to delete your chat data.

5. OR tap Delete account to delete your entire account.

Deleting your account is permanent. DeepSeek's policy says you won't be able to reactivate it or retrieve any of the content tied to it.

Bottom line: Is DeepSeek safe?

DeepSeek is safe enough for low-stakes questions and a poor choice for anything personal. Its privacy policy confirms it collects your prompts, uploaded files, photos, voice input, and chat history, and stores it all in China. Most major LLMs collect a similar amount of data. The difference with DeepSeek is which country's laws govern it.

On functionality, we preferred ChatGPT. We used the free version of each and found DeepSeek to be slower, with less diversity and organization in its answers.

If you use DeepSeek, limit what it gets. Opt out of model training, keep identifying details out of your prompts, and delete your chats when you're done. If there's a teen in your house using AI chatbots, parental controls give you more oversight than DeepSeek's policy does.

FAQs