Is DeepSeek safe to use?

What DeepSeek collects

DeepSeek’s privacy policy, last updated February 10, 2026, covers its apps, website and related services, run by Hangzhou DeepSeek Artificial Intelligence Co., Ltd. With an account it collects your email address or phone number, password and, where applicable, date of birth. It also collects what you put in: text, voice input, prompts, uploaded files, photos, feedback and chat history. The policy says it does not extract voiceprints or facial features from voice or photos.

Every visit adds device and network data: device model, operating system, IP address, device identifiers, system language, crash reports and logs of the features you use. It works out an approximate location from your IP address. If you sign in with Apple or Google, it receives data from that account, and it says it may gather publicly available personal data from the internet to train its models.

The policy says the service is not meant for sensitive data such as health, religion, sexuality or immigration status, and asks you not to provide it, about yourself or anyone else.

Where your chats are stored, and who sees them

The policy states that DeepSeek collects, processes and stores your personal data directly in the People’s Republic of China. Entities in its corporate group process it for storage, security, research, analytics and foundation model training. It also shares data with service providers, including the search services that receive your keywords when DeepSeek looks something up, and with analytics and safety monitoring vendors.

It may share personal data with law enforcement, public authorities or other parties when it believes in good faith that the law, legal process or a government request requires it, and with buyers in a merger or sale. The policy says DeepSeek does not run targeted advertising, sell personal data or build profiles that make significant decisions about you.

Storage in China is the main reason regulators acted. Italy’s data protection authority ordered the app blocked in January 2025. South Korea’s authority suspended new downloads in February 2025 until DeepSeek fixed problems with its data handling. In June 2025, a German data protection commissioner asked Apple and Google to remove the app, saying it transferred users’ data to China unlawfully. Several governments, including South Korea, Australia and Taiwan, barred it from official devices.

How long it keeps your chats

DeepSeek gives no fixed period. Its policy says the data it needs to provide the service, including your account details and what you input, is kept for as long as you have an account. It may keep data longer for legal obligations, legitimate business interests such as improving the service and its security, or to handle a violation of its terms. You can copy or delete your chat history in settings, and deleting your account cannot be undone.

Security has slipped before. In January 2025, the cloud security firm Wiz found a DeepSeek database open to the internet that held more than a million log lines, including chat history in plain text; DeepSeek closed it after being told. Security researchers cited by Krebs on Security also found its iPhone app sending some user and device data without encryption.

When your chats train its models

The policy lists training and improving its models among the uses of your data, and gives you the right to opt out. DeepSeek’s own explanation of its training says a small portion of fine-tuning data may be built from user input, which it says it encrypts, de-identifies and anonymizes first.

To opt out, open Settings, then Data (Data controls in the mobile app), and turn off “Improve the model for everyone.” That covers new chats; for chats you already sent, guides suggest emailing privacy@deepseek.com with the specific conversations.

The app versus the open models

DeepSeek publishes its model weights openly, so anyone can download and run them. A DeepSeek model running on your own computer sends nothing to DeepSeek. A DeepSeek model offered by another company follows that company’s privacy policy, not DeepSeek’s; its storage location, retention and training rules are whatever that host sets.

Its policy also says it applies only to DeepSeek’s own services; apps built on its developer platform have their own rules. So the question to ask is not only which model, but who runs it and where your prompts are stored.

How to limit what DeepSeek keeps

Turn off “Improve the model for everyone.” Delete chats you no longer need, and delete your account when you are done, since your data is kept for as long as it exists. Avoid sharing a chat by public link unless you mean it to be public; the policy warns that shared links posted online can be picked up by crawlers.

Leave out what the task does not need: names, addresses, account numbers, work documents and anything about other people. If your employer or government bars the app, that rule applies whatever settings you choose. For sensitive work, a locally run open model avoids sending your prompts to anyone.

How nokeep differs

nokeep needs no account, so there is no email, phone number or date of birth to collect. Saved chats live in your browser and nokeep keeps no server copy of them; it does not train on your chats and runs no ads or advertising trackers. The tradeoff is that nothing syncs between devices, and clearing browser data erases your history.

nokeep is not open source and does not run on your computer: it uses hosted models, so a prompt still has to be processed to be answered, under terms that the provider does not retain it afterwards. Free use covers 25 messages and 2 images a day, plus one short video clip while the daily free supply lasts.