Is DeepSeek safe? A look at the risks and how to protect your data

DeepSeek has hit the headlines with the release of its AI chatbot app. Said to be comparable to OpenAI’s ChatGPT for a fraction of the price, it recently became the most downloaded free iOS app in the US. DeepSeek’s arrival caught Wall Street off guard, sending tech stocks tumbling. Notably, chip-making company Nvidia lost 17 percent (almost $600 billion) of its market value in a single day in January 2025, which was the biggest single-day loss in US stock market history.

The overnight success of DeepSeek is all good, but the key question many people find themselves asking is whether DeepSeek is safe to use. The biggest concern, of course, is whether your data is being harvested and privacy compromised in the process, with this being a Chinese chatbot app. In this post, we’ll explain what DeepSeek is, the kind of data DeepSeek collects, some of our concerns, and whether you can use it safely.

What is DeepSeek?

DeepSeek is a Chinese artificial intelligence company. Founded in 2023 by Liang Wenfeng, it develops large language models (LLMs), most notably DeepSeek V3 and DeepSeek-R1.

Some AI enthusiasts compare the quality of DeepSeek’s responses with ChatGPT’s. That’s no mean feat if DeepSpeak’s claim that it cost just USD $6 million to train its flagship AI model compared to the $100 million of ChatGPT’s latest model. However, this figure has since been contested by a report from SemiAnalysis that estimated DeepSeek’s hardware spend to be over $500 million.

Like rival AI chatbots ChatGPT and Gemini, DeepSeek can answer questions, generate content and code, and even solve complex problems. The key difference is that, unlike other tools that require you to fork out a paid subscription to enjoy their full benefits, DeepSeek is entirely free, at least for now.

You can choose from two DeepSeek LLMs: DeepSeek-V3 and DeepThink (R1), both of which are accessible via web browser and Android and iOS apps. V3 works much like the Standard ChatGPT model and provides fast responses, while R1 is an alternative to the ChatGPT o1 model and doesn’t require a paid subscription.

Why is everyone talking about DeepSeek?

The launch of DeepSeek’s R1 model has made significant waves in the AI industry. Until recently, the US has dominated the AI space, but now a new Chinese competitor has emerged, which claims to be using far less powerful computer chips. DeepSeek is also open-source and free to use, which is a big reason for its explosion in popularity. Consumers can thoroughly test it out and explore its capabilities without limits.

What data does DeepSeek collect?

You probably won’t be surprised to find that DeepSeek collects a lot of data. Still, knowing exactly what data DeepSeek collects helps you decide whether using it is worth the risk.

DeepSeek’s privacy policy details three ways in which it collects your information:

1. Information you voluntarily provide, such as during signup, when inputting content such as text or audio, or when you get in touch with DeepSeek.
2. Data that’s automatically collected when you use DeepSeek’s services, such as device identifiers.
3. Information from linked services and advertising partners.

Here’s some of the data that DeepSeek collects:

Profile information

Although DeepSeek is free, you must sign up for an account. This requires providing an email address or a phone number. DeepSeek keeps track of this, as well as the password you create. Its privacy policy also mentions logging your date of birth, but we weren’t asked to provide this during signup.

User input

Everything you type or upload to DeepSeek is logged. This includes all text and audio inputs, meaning that, unsurprisingly, none of your chats or prompts are private.

Technical data

DeepSeek collects device information, including your device model, operating system, system language, and keystroke patterns. It also records your IP address (and thus your approximate location).

Usage data

DeepSeek collects information that helps it understand how you’re using its service, such as the features you’re using and specific actions you’re taking.

Cookies and trackers

The privacy policy of DeepSeek mentions not only cookies but “other similar technologies,” referring to “web beacons.” These are trackers that follow how you’re using DeepSeek’s services.

Third-party data

The data DeepSeek collects isn’t limited to its own platform and services. Through various partners, it receives information about you and your online activities via other websites, apps, and stores. Information shared with DeepSeek may include mobile identifiers, hashed email addresses, and phone numbers.

How does DeepSeek use your data?

Between collecting all of your chats, IP address, unique device identifiers, and data from third parties, DeepSeek can learn a lot about its users, building substantial profiles of each. The question is, what does it do with all of this data?

DeepSeek states that the information it collects helps it “operate, provide, develop, and improve” its service. More specifically, it explains that logging chats with the AI chatbot is to enforce its terms and ensure the safety of its users.

DeepSeek may use collected data (such as your email address) to notify you about any changes to the service. DeepSeek’s privacy policy also cites the identification and addressing of bugs as well as the improvement of its service and technology as further reasons for data logging.

The privacy policy’s mention of complying with legal obligations is expected but particularly relevant. DeepSeek is a Chinese app subject to Chinese laws requiring it to provide the Chinese government with data upon request.

All of the above is what DeepSeek itself covers as part of its privacy policy. Privacy advocates worry that DeepSeek can build up detailed profiles of users and use them for highly targeted advertising or even to influence a person’s views, such as those related to politics.

Who does DeepSeek share your data with?

Much of the data that DeepSeek collects is shared. Here’s who it’s shared with:

Service providers

Information you willingly provide that’s automatically collected from other sources, such as linked services and advertising and measurement, can be shared with service providers. For example, if you sign up or log in via a third-party platform such as Google, DeepSeek will provide Google with specific technical information to make this possible.

Business partners

DeepSeek shares information about your use of its service (the features you use and actions you take) with advertising or analytics partners. Notably, these partners help DeepSeek “promote the Service through advertising displayed on other platforms, some of which, in some jurisdictions, may be based on your activity across other sites and services.”

This suggests that cross-platform tracking through cookies and device identifiers sees third-party advertising partners using your online activities (such as the websites you visit) to provide targeted ads. By DeepSeek AI chatbot’s own words, this “may raise privacy concerns.”

DeepSeek’s corporate group

It’s vague, but DeepSeek’s privacy policy refers to “certain entities within our corporate group.”

These entities are said to perform the following tasks:

• Storage
• Content delivery
• Security
• Research and development
• Analytics
• Customer and technical support
• Content moderation

The information shared can be anything from your chats with the AI chatbot and any files you share to your email address, IP address, or keystroke patterns.

Chinese authorities

DeepSeek mentions “others” with regard to who it may share your data with. As part of this, it references government requests, which is, of course, one of the big privacy concerns surrounding DeepSeek. DeepSeek is a Chinese company that stores information collected on servers located in China.

Considering the Chinese government has significant authority to access user data, DeepSeek has little alternative but to hand over user data. Due to China’s extensive surveillance and censorship, this data may be used to monitor individuals or even to persecute minority groups, journalists, or activists.

Related: China VPNs

Is DeepSeek safe? What are the risks?

DeepSeek’s data collection is concerning. You can expect much of the same logging from other popular AI chatbots such as ChatGPT and Gemini. However, the key difference is that Western companies such as OpenAI (ChatGPT) and Google (Gemini) have more legal avenues to explore when resisting data requests from authorities.

With DeepSeek operating under China’s National Intelligence Law, it has little choice but to comply with data requests, regardless of the impact on user privacy. DeepSeek’s logging of keystrokes takes things a step further than most AI chatbots. Even if you type a message to the chatbot and delete it before sending it, DeepSeek can still record the input.

The fact that both Ireland’s Data Protection Commission (DPC) and Italy’s Data Protection Authority (DPA) are investigating how DeepSeek collects and stores data is telling. Indeed, Italy’s DPA has blocked access to DeepSeek in Italy due to the company’s failure to provide enough information regarding its data handling practices.

Just as concerning as DeepSeek’s data logging is its security practices, mainly after Wiz Research found a publicly accessible DeepSeek database leaking over a million lines of data. This included sensitive information such as chat history and secret keys. Though DeepSeek promptly addressed this vulnerability upon being notified by Wiz, it’s enough to question DeepSeek’s security practices.

There’s also no shortage of evidence of censorship on DeepSeek. Ask DeepSeek to tell a political joke about a world leader such as Donald Trump, and it duly obliges. However, it then declines to do the same for China’s leader, Xi Jinping. This makes it all the more difficult to trust an AI tool that may also provide misinformation, unintentionally or otherwise.

Despite the numerous negatives, DeepSeek is at least open source, meaning anyone can inspect code and improve it. Although we may be keen to adopt AI tools such as DeepSeek and the convenience they provide, the fact remains that entrusting them with a lot of data, much of it of a sensitive nature. Regardless of how fast AI is evolving, companies like DeepSeek that offer services must provide sufficient security to protect our data.

What can I do to protect my data on DeepSeek?

Although DeepSeek’s data collection and security practices are concerning, there are some things you can do to protect your privacy better when using its services:

Read DeepSeek’s privacy policy

We’ve already gone over some of DeepSeek’s privacy policies and the data it collects. Still, it’s worth reading through. The good news is that the privacy policy isn’t overly long or difficult to read, and the information there might help you weigh up the risk of using DeepSeek versus seeking an alternative.

Avoid sharing personal or sensitive data

What’s clear is that you shouldn’t enter any personal data, particularly data of a sensitive nature, with an AI chatbot such as DeepSeek. With the extensive data collection involved, this information will be stored, studied, and even shared with third parties such as the Chinese government. A data breach could also mean your sensitive information is compromised.

Use a privacy-focused browser and a VPN

Due to the level of tracking on DeepSeek, it’s worth using a privacy-centric browser such as Brave. This blocks tracking scripts from running, making it harder for DeepSeek to follow you across the web. Brave also offers the option to browse with Tor. This routes traffic via the Tor network, adding another anonymity layer.

The best VPNs can further protect you using DeepSeek by encrypting your connection and hiding your IP address and location. Of course, this should be a no-logs VPN for genuinely anonymous browsing.

Our top recommendation is NordVPN, a market-leading VPN renowned for its security and privacy features. Comparitech readers can get an exclusive discount by following the link below.

Is DeepSeek safe? FAQs