DeepSeek ban: US orders agencies to remove Chinese AI; Beijing objects
Representative image: Mariakray / Pixabay
The US Homeland Security department ordered civilian federal agencies to remove DeepSeek AI from their systems, made public on October 19, 2026; China says it will protect its firms.
What Happened
The DeepSeek ban in the United States widened this week. The US Department of Homeland Security (DHS) has ordered civilian federal agencies to remove technology from DeepSeek, the Chinese artificial intelligence company, from their computer systems, in an order made public on Monday, October 19, 2026. The order also bars those agencies from buying DeepSeek products, including third-party services built on its AI models. China's foreign and commerce ministries objected on October 20 and 21, saying Beijing would protect the rights of Chinese companies.
Key Facts
- DHS ordered US civilian federal agencies to remove DeepSeek technology from their IT systems; the order was made public on October 19, 2026.
- It covers DeepSeek apps and also third-party products and services built using its AI models, and it bars agencies from buying them.
- The Department of Defense is not covered by this order; the Commerce Department and the Navy had already banned DeepSeek on their devices.
- DHS said prompts and answers in some DeepSeek apps could lead to the "transfer of sensitive data to the PRC" and that the models may contain malware or security weaknesses.
- Foreign ministry spokesperson Mao Ning said China opposes "the overstretching of the concept of national security" and will safeguard Chinese firms' rights.
- China's commerce ministry urged Washington to revoke what it called discriminatory restrictions, saying the move ignored understandings reached by the two countries' leaders.
Why It Matters
DeepSeek is known for releasing "open-weight" AI models, meaning anyone can download and run them on their own computers; its latest such release came earlier this month. That openness is part of why the new order reaches further than earlier bans. Previous restrictions mostly stopped staff from installing the DeepSeek app on government phones. This order also covers other companies' software that is built on DeepSeek's models. A US contractor that quietly uses a DeepSeek model inside a product sold to an agency would now have to remove it.
The stated worry is about data and hidden flaws. When someone types a question into a chatbot, the question and the answer travel to the company's servers. DHS argues that for DeepSeek's apps those servers are in China, where authorities could seek access. It also says the models themselves might hide malware, harmful code or weak points. DeepSeek has not been shown in the reports to have misused government data; the order is a precaution based on where the company is based and how its products work.
For China, the move is another sign that the US treats AI as a security issue, not just a trade one. Beijing's response linked it to recent leaders' talks between the two countries and to US sanctions on Chinese entities over alleged AI trade-secret theft. Several US states, including New York, Tennessee and Virginia, have already taken similar steps for their own devices.
For Indian start-ups and developers, the case is a reminder that the choice of AI model can decide which customers a product can sell to. Many low-cost Indian AI products are built on open models from around the world. A product that relies on a model from a country under suspicion may find government and large company buyers closed to it. Choosing models with clear origin, hosting and data rules is becoming a business decision, not only a technical one.
| DeepSeek order at a glance | Detail |
|---|---|
| Issued by | US Department of Homeland Security |
| Made public | October 19, 2026 |
| Covers | Civilian federal agency IT systems, including products built on DeepSeek models |
| Not covered | Department of Defense |
| China's response | Foreign and commerce ministries objected, October 20–21, 2026 |
Impact
Short-term: US civilian agencies and their contractors must find and remove DeepSeek apps and any products that use its models, and stop buying them. Software firms selling to the US government will need to check what AI models sit inside their products.
Long-term: The order deepens the split between US and Chinese AI ecosystems. Other governments may follow with similar rules, and buyers may increasingly ask suppliers to declare which AI models they use and where data is processed.
Who is affected: US federal civilian agencies and their contractors, DeepSeek and companies building on its models, Chinese technology firms more broadly, and start-ups worldwide, including in India, that sell AI products to governments or large enterprises.
Key Takeaway
The US DeepSeek ban now reaches beyond the app to any federal product built on its AI models, turning the choice of model into a compliance question for AI companies.
Questions and Answers
What does the US DeepSeek ban cover?
A Department of Homeland Security order, made public on October 19, 2026, requires US civilian federal agencies to remove DeepSeek technology, including third-party products built on its AI models, from their IT systems and stop buying them. The Defense Department is not covered by this order.
Why did the US ban DeepSeek from federal systems?
DHS said the way DeepSeek's apps handle prompts and answers could lead to the transfer of sensitive data to China, and that its models may contain malware or security weaknesses.
How did China respond to the DeepSeek ban?
Foreign ministry spokesperson Mao Ning said China opposes the overstretching of national security and will safeguard Chinese companies' rights. The commerce ministry urged the US to revoke the restrictions.
Does the DeepSeek ban apply to ordinary users?
No. The order applies to US civilian federal agencies' systems and purchases, not to private individuals or companies, though contractors selling to those agencies are affected.
Disclaimer: Prepared by the Peepals newsroom from publicly available sources with AI assistance. Information is accurate to the best of our knowledge at the time of publication and may change. Images may be representative. Not professional advice. Report an error via our contact page.
Sourced and fact-checked by the Peepals Global Editorial Team
eMudhra certificate keys compromised: browsers block eight, Chrome to drop a root
ChatGPT adds Locked Chats with PIN and Face ID, but no encryption
Reported, fact-checked and published by the Peepals Global Editorial Team.
