If you’re a ChatGPT user, you might have recently tried DeepSeek. The Chinese AI chatbot seems so promising that it’s starting to challenge many U.S. tech giants. DeepSeek is an open source solution that can perform exceptionally well locally (on certain hardware). It also offers a free online version that currently allows unlimited queries. Additionally, there’s a paid API with very competitive rates.
When you use chatbots like DeepSeek online, you often share a substantial amount of information. These tools have become valuable allies for tasks such as planning vacations, summarizing documents, creating budgets, and analyzing images. Their language models process every word you input to provide you with answers or approximations that you can refine with further prompts.
What Does DeepSeek Do With Your Data?
After the initial excitement of discovering a new AI tool, questions may arise about what exactly happens to your data. Does the data vanish after being processed by the model? Are you unknowingly granting permanent access to a company? Is it stored as a valuable resource to train future versions of the model?
These concerns aren’t entirely new. Many, including European regulators, raised similar questions when ChatGPT gained popularity. Now, with DeepSeek emerging as the hot new AI model, these questions are once again on the horizon.
Amid the uncertainty, one thing is clear: DeepSeek’s web version and mobile app collect a substantial amount of data. This may not surprise some users, but it could be a revelation for those just starting to use the chatbot. For more information about the Chinese company’s data privacy practices, visit its privacy policy page.
Hangzhou DeepSeek Artificial Intelligence Co., Ltd. and Beijing DeepSeek Artificial Intelligence Co., Ltd. gather users’ profile information. This includes usernames, dates of birth (if applicable), email addresses, phone numbers, and passwords. Additionally, these companies collect chats, encompassing texts, audio files, uploaded documents, comments, and chat history.
If you have a question about DeepSeek and decide to reach out through its contact channels, the companies behind DeepSeek will collect all the information you provide. This includes proof of identity or age, as well as any comments or queries you have about the service. This type of information is referred to as “information you provide.”
In addition to the information you provide, there’s another category of data called “automatically collected information.” This includes details such as your device model, operating system (including its language), IP address, cookies, and diagnostic and performance data. The companies may also track your keystroke patterns. All this data will be linked to a device ID and user ID.
You can access the web version of DeepSeek at chat.deepseek.com.
While it’s difficult to assign a monetary value to your data, it’s undoubtedly valuable. One method of assessing its worth is through the various outcomes it generates. First and foremost, DeepSeek uses the collected data to train its AI models. The companies also say they monitor interactions, although it remains unclear whether human analysts are reviewing conversations at this stage.
The privacy policy page reveals further significant information. It says, “We review User Input, Output and other information to protect the safety and well-being of our community.” The companies also collect data to comply with legal obligations, “perform tasks in the public interest,” and notify users about changes to services.
Data Collected by DeepSeek Doesn’t Remain Solely With the Company
According to its privacy policy, DeepSeek may share information gathered from various categories with others. This includes “certain entities within our corporate group,” which means actors who are part of the companies behind DeepSeek. Additionally, DeepSeek could share information “advertising or analytics partners.”
As previously mentioned, the collected information is tagged with a device ID and a user ID. These identifiers can be quite useful for tracking user activity and cross-referencing it with data from other platforms. While DeepSeek mentions that it may use activity from other sites and services, this is subject to regulations in certain jurisdictions.
Moreover, DeepSeek says that it may share collected information with “law enforcement agencies, public authorities, copyright holders, or other third parties if we have good faith belief that it is necessary to.” This means that the company can provide data to government entities. While this practice is common in many jurisdictions, it’s particularly concerning in China, where there have been multiple controversies surrounding data sharing.
Numerous investigations have scrutinized the Chinese Communist Party and the laws of the People’s Republic of China, which compel technology companies in the country to provide data to authorities. A Department of Homeland Security document indicates that the U.S. government encourages companies to install backdoors to facilitate operations aimed at maintaining national security.
One of TikTok’s main issues was the exposure of American users’ data to foreign entities. To alleviate this concern, ByteDance, the social media platform’s parent company, reached an agreement with Oracle to store U.S. user data within the U.S., subjecting it to U.S. law.
While users can delete DeepSeek’s chat history, it’s not specified whether this data is also removed from the servers, which would prevent it from being used for training purposes. Additionally, there doesn’t seem to be an option to prevent chat content from being utilized to enhance future iterations of the underlying language model.
The companies’ private policy addresses some of these concerns. For instance, it states that registered users can “access, review, and update certain personal information.” However, the process for doing so doesn’t appear to be straightforward. DeepSeek encourages users who wish to exercise their rights to contact them directly or manage some data collection preferences through their browser’s cookie settings.
Image | DeepSeek | Alejandro Luengo
View 0 comments