is deepseek safe is deepseek safe to use is deepseek safe to run locally

Is DeepSeek safe? What its privacy policy actually says

Adrian Maverick · · 8 min read

DeepSeek is a real AI company with a working chatbot, not malware, but its own privacy policy is clear that what you type is stored in China and can be used to train its models. Whether it is safe for you depends on what you put into it and whether you are comfortable with those two facts.

Here is what DeepSeek collects, where it keeps it, how to switch off training, and why some governments have restricted it, taken from DeepSeek's own policy pages and from official government statements, checked on November 15, 2026.

Who runs DeepSeek and where your data goes

DeepSeek's privacy policy, last updated February 10, 2026, names the company in control of your data as Hangzhou DeepSeek Artificial Intelligence Co., Ltd., "with its registered address in China."

On storage it is direct: "To provide you with our services, we directly collect, process and store your Personal Data in People's Republic of China." That applies wherever you live.

The terms of use, last updated March 27, 2026, add that disputes are governed by the laws of mainland China and go to a court near the company's registered office in Hangzhou.

The privacy policy also says DeepSeek may share data with law enforcement and public authorities to comply with "applicable law, legal process or government requests," in accordance with "the due process prescribed by law." In practice, that means the laws that apply to your chats are Chinese law, not the privacy law of your own country alone.

What DeepSeek collects

The privacy policy lists these categories:

  • Account details. Date of birth where applicable, username where applicable, email address and/or phone number, and password.
  • Everything you put in. Text and voice input, prompts, uploaded files, photos, feedback and chat history. DeepSeek says it will not extract voiceprints or facial recognition data from your voice or photos.
  • Device and network data. Device model, operating system, IP address, device identifiers, system language, crash reports and performance logs. DeepSeek assigns you a device ID and user ID and uses them to link your activity across devices.
  • Usage logs. The features you use and the actions you take.
  • Approximate location. Worked out from your IP address, used for security and for location questions like the weather in your city.

Two more details are worth knowing. When you use search, DeepSeek says it shares "your input keywords" with third-party search providers. And if you share a chat by link and it ends up public, DeepSeek warns it can be picked up by web crawlers.

On the positive side, the policy says DeepSeek does not do targeted advertising and does not sell personal data.

Does DeepSeek train on your chats?

It can, unless you opt out. The privacy policy says DeepSeek uses your data "to train and improve our technology, such as our machine learning models and algorithms," and that companies in its corporate group handle "foundation model training and optimization."

DeepSeek's model and training explainer says most of its fine-tuning data is written by its research team, with "a small portion potentially based on user input." It says user input used this way is encrypted and de-identified first.

How to turn off training in DeepSeek

DeepSeek's terms name the setting: you can opt out "by turning off 'Improve the model for everyone'."

  1. Sign in to DeepSeek on the app or website.
  2. Open your settings.
  3. Find Improve the model for everyone and switch it off.

DeepSeek's pages do not describe the exact menu path, so look for that name in your settings. The privacy policy also lists a "right to opt-out of using your Personal Data for training our models," and you can email privacy@deepseek.com if you cannot find the switch.

For the same setting in other AI tools, see how to opt out of AI training.

How to delete your DeepSeek data

  • Chat history. The privacy policy says you can manage, copy or delete your chat history in your settings.
  • Your account. You can delete your account. DeepSeek says you then cannot reactivate it or get back any of its content.
  • What may stay. The terms say that even after you delete your account, DeepSeek may "retain certain data of the user as required by laws and regulations." The policy gives no specific retention period.
  • Formal requests. Access, correction, deletion and portability requests go to privacy@deepseek.com. In the EU and UK, DeepSeek's privacy representative is Prighter, at rep_deepseek@prighter.com.

Why some governments restrict DeepSeek

You may have heard that DeepSeek is "banned." The official actions we checked are narrower than that word suggests.

  • New York State. On February 10, 2025, Governor Hochul banned downloading DeepSeek on state government devices and networks managed by the state's IT office, citing concerns about "foreign government surveillance and censorship." It is a rule for government devices, not for people's personal phones.
  • Italy. On January 30, 2025, Italy's data protection authority, the Garante, ordered an urgent limitation on DeepSeek processing Italian users' data. The Garante said the companies told it they do not operate in Italy and that European law does not apply to them, called that reply "entirely unsatisfactory," and opened an investigation.

These actions are about where data goes and which laws apply, not proof that anything bad happened to a particular user. If you work for a government agency or a company with an AI policy, those rules come first.

Is DeepSeek safe to run locally?

This is a different question from using the app. DeepSeek says it releases its model weights, parameters and inference code under the MIT License, so anyone can download and run the models on their own hardware.

DeepSeek's privacy policy covers its own apps, websites and services. A model running on your own computer does not send your prompts to DeepSeek's servers by itself, so the storage and training points above do not apply. Check the software you use to run it, since that is what handles your data.

Apps built by other developers on top of DeepSeek's platform are a third case. DeepSeek's policy says it does not cover them, and the developer's own privacy policy applies.

Practical rules if you use DeepSeek

  • Switch off Improve the model for everyone before you start.
  • Do not type anything you would not want stored in China: passwords, ID numbers, bank details, health information, or other people's private messages. DeepSeek's own policy says not to share sensitive personal data.
  • Do not paste in work documents unless your employer allows it.
  • Be careful with shared chat links.
  • Delete old chats you no longer need.
  • Do not treat answers as professional advice. DeepSeek's terms say outputs may contain errors and are for reference only.

Frequently asked questions

Is DeepSeek safe to use in the USA?

There is no single answer for every user. The official action we checked, New York's, covers state government devices and networks. For a personal account, the main facts are the ones in DeepSeek's own policy: your data is stored in China, and it can be used for training unless you opt out.

Does DeepSeek train on my data?

Yes, by default it can. DeepSeek's terms say it may use de-identified inputs and outputs to improve its services, and that you can opt out by turning off Improve the model for everyone.

Is DeepSeek safer than ChatGPT?

Compare the policies, not the brands. The biggest difference in DeepSeek's case is that its policy says data is stored in China and governed by Chinese law. For other assistants, see is Claude private, is Copilot private and is Gemini private.

Is DeepSeek safe to download on iPhone?

Download it only from the official app store listing. The privacy questions above matter more than the download itself.

Does DeepSeek sell my data?

Its privacy policy says it does not sell personal data and does not do targeted advertising.

Keep sensitive details out of AI chats

The safest habit with any AI chatbot is the same: keep passwords, account numbers and private messages out of it. A lot of that information lives in your inbox, so it is worth knowing what your email provider does with it too.

Secria Mail has no ads, and Secria never mines your mail. Your mailbox is locked with a key made from your password, on your device. Sign-up is a name, the address you want and a password, with no phone number. Create a free account.

Secria fact-checks every post against primary sources. Spotted something wrong or out of date? Email hq@secria.me and we will correct it.