DeepSeek has quickly become one of the most closely watched names in generative AI. Its rise is not only about model performance or competition with US platforms. It is also about what happens to user data when millions of people try a new AI service through an app, a website or a chat interface.
The company’s own privacy language is direct about where collected information is stored. For users in the United States and elsewhere, that makes DeepSeek a case study in how AI convenience, data collection and cross-border storage now meet in everyday software.
A fast rise brings privacy questions
DeepSeek is an AI research lab created by a prominent Chinese hedge fund. It recently gained attention after releasing its latest open source generative AI model, which the source article says easily competes with top US platforms like those developed by OpenAI.
The timing matters. Recent US regulatory action against TikTok pushed many users toward another Chinese app, Rednote. DeepSeek’s popularity now offers another example that restrictions on one Chinese-owned digital service do not necessarily stop Americans from using others.
DeepSeek has several AI models. Some can be downloaded and run locally on a laptop. But most people are likely to use the service through iOS or Android apps or the web chat interface, where they can ask questions, get answers, search the web or use a reasoning model to elaborate on responses.
On Monday, DeepSeek’s creators limited new sign-ups after saying the app had been overrun with a "large-scale malicious attack." DeepSeek did not return WIRED’s request for comment about user data protections and privacy priorities.
What DeepSeek says it collects
The key privacy issue is not hidden in vague language. The English-language DeepSeek privacy policy says: "We store the information we collect in secure servers located in the People's Republic of China."
That means conversations, questions and the answers generated by the service are being sent to China or can be. The policy describes three broad categories of information: data users provide, data collected automatically and data received from other sources.
The first category includes user input. According to the policy, DeepSeek may collect text or audio input, prompts, uploaded files, feedback, chat history and other content provided to the model and services.
That is similar in shape to how other generative AI platforms operate, because prompts are central to how these tools answer questions. But privacy advocates cited in the source article stress a practical rule: users should avoid putting sensitive or personal information into AI chat bots.
Lukasz Olejnik, independent researcher and consultant, affiliated with King's College London Institute for AI, put the advice plainly: "I would not input personal or private data in any such an AI assistant."
DeepSeek also collects account-related information. The source article lists email address, phone number, date of birth, username and more. If a user contacts the company, that communication can also share information with DeepSeek.
Automatic data collection expands the picture
DeepSeek’s privacy picture is not limited to what a person types into a chat box. Like many digital platforms, it can collect information automatically while people use the service.
The source article says DeepSeek can collect device information, operating system details, IP address and crash reports. It can also record "keystroke patterns or rhythms," a kind of data more widely collected in software built for character-based languages.
If a user purchases DeepSeek’s premium services, the platform collects that information too. It also uses cookies and other tracking technology to "measure and analyze how you use our services."
WIRED’s review of the DeepSeek website’s underlying activity found that the company also appears to send data to Baidu Tongji, Baidu’s web analytics tool, and Volces, a Chinese cloud infrastructure firm.
Sean O'Brien, founder of Yale Law School's Privacy Lab, said in a social media post that DeepSeek is also sending "basic" network data and "device profile" to TikTok owner ByteDance "and its intermediaries.
Data from other sources and AI’s tradeoff
DeepSeek’s policy also reserves the right to collect information from other sources. If someone creates an account using Google or Apple sign-on, DeepSeek receives some information from those companies.
Advertisers may also share information with DeepSeek, according to its policies. The source article says that can include "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."
For users, the broader issue is that generative AI systems can feel conversational and personal while operating under company-defined terms. John Scott-Railton, a senior researcher at the University of Toronto’s Citizen Lab, warned that the concern is not limited to one company: "It shouldn’t take a panic over Chinese AI to remind people that most companies in the business set the terms for how they use your private data" and "when you use their services, you’re doing work for them, not the other way around."
Bart Willemsen, a VP analyst focusing on international privacy at Gartner, said the construction and operation of generative AI models is generally not transparent to consumers and other groups. People do not know exactly how the systems work or the precise data they were built on.
DeepSeek is largely free for individuals, though developers pay costs when using its APIs. Willemsen framed the exchange this way: "So what do we pay with? What do we usually pay with: data, knowledge, content, information."
What users can do with that information
The source article identifies a few practical choices. Users can delete chat history inside DeepSeek’s settings. On mobile, that means using the left-hand navigation bar, tapping the account name at the bottom of the menu to open settings and choosing "Delete all chats."
Another difference is where the model runs. Olejnik notes that when models like DeepSeek’s are installed locally and run on a user’s computer, people can interact with them privately without sending data to the company that made them.
Perplexity says it has added DeepSeek to its platforms but claims it is hosting the model in US and EU data centers. That distinction matters because the same underlying model can be offered through different technical arrangements.
DeepSeek’s rise shows why AI privacy cannot be treated as an afterthought. The tool may be powerful, popular and easy to try, but the data behind each chat still has a destination, a policy and a business use.