The United States government just ordered one of the world's most talked-about AI chatbots ripped out of federal computers — and the order does not stop at the app with DeepSeek's name on it. On Monday, October 19, 2026, the Department of Homeland Security made public an exclusion and removal order directing every federal civilian agency to stop buying DeepSeek technology and to remove it from its systems, including any third-party product built on top of DeepSeek's models. The filing, scheduled for formal publication in the Federal Register on Tuesday, October 20, is the first time Washington has used its supply-chain exclusion power against a named AI company at this scale.
There is a twist that makes the story stranger. The order itself — Exclusion and Removal Order #2026-01 — was signed back on July 24, 2026. It only became public this week, almost three months later. And the window agencies were given to request an exception, 60 days from issuance with a compelling written justification, had already closed before most people outside government ever saw the document.
Key Takeaways
- DHS Exclusion and Removal Order #2026-01, signed July 24, 2026 and filed publicly on October 19, orders all federal civilian agencies to halt purchases of DeepSeek products and remove existing DeepSeek software from their IT systems.
- The ban covers more than the DeepSeek app: third-party products and services built using DeepSeek implementations must go too, which means agencies have to audit what is inside their vendors' software, not just its brand name.
- DHS says DeepSeek's models "present multiple threat vectors," including the risk that sensitive data could be transferred to the People's Republic of China, that the models could be used to track or fingerprint users and build intelligence on Americans, and that the models themselves may contain malware or security vulnerabilities.
- The order applies to civilian executive branch agencies — the Department of Defense is not covered — and it does NOT ban private Americans or private companies from using DeepSeek. Federal Register document 2026-21447, docket DHS-2026-0628, is scheduled for publication on October 20, 2026.
What the DHS Order Actually Does
The order names Hangzhou DeepSeek Artificial Intelligence Co., Ltd., the Chinese company behind the DeepSeek chatbot and its open-weight models. It was issued by the Secretary of Homeland Security under the Federal Acquisition Supply Chain Security Act of 2018, after a review and recommendation by the Federal Acquisition Security Council — the interagency body that decides which products are too risky for federal systems.
It does two things at once. First, exclusion: civilian agencies may not buy DeepSeek products or services, or those of its affiliated entities, going forward. Second, removal: agencies must take DeepSeek software already inside their systems out — and the removal duty reaches "third-party products and services developed using DeepSeek implementations." Exception requests go through the Cybersecurity and Infrastructure Security Agency (CISA), part of DHS, and had to be filed in writing within 60 days of the July 24 issuance, with a compelling justification and proposed safeguards.
The order is explicitly limited. It covers civilian executive branch agencies — departments like Treasury, State and Justice — and not the Department of Defense. Reporting on the filing also notes carve-outs for national security systems. And it is a procurement and government-systems order, not a consumer ban: it does not make it illegal for a private person or a private business in the US to download or use DeepSeek.
Why Washington Says DeepSeek Is a Risk
DHS's language, quoted across Monday's reporting by Reuters, the Wall Street Journal, Bloomberg, CyberScoop, Quartz and SiliconANGLE, is unusually blunt for a procurement document. The department said DeepSeek's models present "multiple threat vectors" for federal systems. Use of the products by government employees or vendors, it warned, "may result in the transfer of sensitive data to the PRC," and the models could "produce actionable intelligence on US persons and/or government agencies." Other summaries of the order add that the technology could be used to track or digitally fingerprint users, that Chinese law and the Chinese Communist Party could give Beijing extensive authority over data the company holds, and that the models themselves may contain malware or other security vulnerabilities.
To be precise about what has and has not been shown: the public notice lays out risks and capabilities, and the reporting on it did not point to a specific confirmed incident in which DeepSeek sent a federal employee's data to China. The government's argument is a supply-chain one — that a tool this deeply embedded in daily work, reading prompts that can contain sensitive internal material, should not be controlled by a company subject to a rival government's data laws. You do not have to agree with the policy to understand the logic: prompts are data, and whoever runs the model sees the prompts.
The Part Most People Missed: Apps Built on DeepSeek Must Go Too
The deepest part of this order is not the famous chatbot. It is the sentence about third-party products. DeepSeek's models are open-weight, which means thousands of developers around the world have built their own apps, plug-ins and business tools on top of them — often without DeepSeek's name appearing anywhere a customer would see it. Under this order, a federal agency cannot keep a vendor's tool just because the box says something else. If the tool was developed using a DeepSeek implementation, it is covered.
That turns a simple uninstall into an audit. Agencies — and the contractors who sell to them — now have to answer a question many software buyers have never had to ask: which model is actually inside this product, and where did that model come from? Commentators compared the structure to the Huawei hardware restrictions, where a federal ban on one company's equipment eventually reshaped buying decisions far beyond the government itself. Whether that happens again is the thing to watch. Just last week, this site covered DeepSeek's newest release, DeepSeek V4.1-Flash-1017, which topped a major coding benchmark at a fraction of rivals' prices. The same week that model was winning developer praise, the US government was publishing an order to remove its family of models from federal systems. Capability and trust, in 2026, are moving in opposite directions.
How We Got Here: From the R1 Shock to a Federal Ban
DeepSeek has been on Washington's radar since it stunned the industry in early 2025 with its R1 reasoning model — a system that matched leading American models from OpenAI and Anthropic on several tests despite being trained more cheaply and on less powerful chips. Its free app was downloaded by millions of Americans, and the privacy questions started immediately; one cybersecurity expert quoted at the time called DeepSeek's privacy policy "not worth the paper it is written on."
The restrictions then arrived in layers. In 2025, the Commerce Department barred DeepSeek from government devices, the US Navy directed its personnel not to use the model over security concerns, and New York State banned it from government devices and networks, citing data privacy and censorship risks. Several states took similar steps for their own employees. Monday's DHS order is broader than all of those: instead of one department's device rule, it is a government-wide civilian exclusion and removal order, issued under a 2018 supply-chain law that was originally designed mainly with hardware in mind — now applied, at full strength, to AI software and model lineage.
What This Means for Regular Americans
If you are a private user in the US, this order does not change what you are allowed to install on your own phone today. Nobody is coming for the DeepSeek app on a personal Samsung Galaxy, Google Pixel or iPhone, and DHS did not announce any consumer action. What it does change is the world around you. Federal employees and contractors should expect DeepSeek and DeepSeek-based tools to disappear from work systems. Companies that sell software to the government will have to document their model supply chain. And state governments, universities and large employers that follow federal signals may adopt similar rules of their own, as several did after the 2025 device bans.
There is also a personal lesson in the government's reasoning that applies no matter which AI app you prefer. Every chatbot prompt you type — a draft contract, a medical question, a photo, a financial document — is data handed to whoever runs that model, under that company's privacy policy and that country's laws. The federal government just decided that trade-off is unacceptable for its own workers. Ordinary users get to make their own call, but it should be a conscious one.
A Practical Checklist for US Readers
- Federal workers and contractors: do not install or use DeepSeek — or a tool you know is built on it — on any work system, and follow your agency's removal guidance. If a vendor tool might contain a DeepSeek model, ask the vendor in writing; that question is now a normal part of federal buying.
- Everyone else: before pasting sensitive material into ANY AI chatbot, check where the company is based, what its privacy policy says it keeps, and whether prompts are used for training. Assume anything you type could be stored on a server you do not control.
- Keep material that should never leave your phone — ID scans, tax and bank documents, private photos, videos and notes — in local, on-device storage such as Pro Calculator Secure Vault, which offers a secure vault for photos, videos, audio, documents and private notes, sign-in protection, and no data collected or shared with third parties.
- If an AI app's headline feature is being free and surprisingly powerful, read that as a prompt to check the privacy policy twice — and be extra careful with lookalike clone apps trading on DeepSeek's name in app stores.
Scam Warning: Fake "DHS Removal" Notices Will Follow This News
Big government tech bans reliably produce phishing waves within days, and this one is aimed at exactly the people least able to ignore an official-looking message: federal employees, contractors and IT staff. Be suspicious of any email or text claiming you must click a link to "certify DeepSeek removal," "verify compliance" with Order #2026-01, or download a "removal tool." DHS and CISA publish orders through the Federal Register and regulations.gov (docket DHS-2026-0628) and communicate with agencies through official channels — they do not send download links or collect credentials by email. If a message pressures you to act immediately, do not click; go directly to your agency's IT security team or the official Federal Register listing and check there.
Frequently Asked Questions
Is DeepSeek now banned in the United States?
No. DHS Exclusion and Removal Order #2026-01 applies to federal civilian executive branch agencies: they cannot buy DeepSeek products and must remove DeepSeek software, including third-party products built on DeepSeek implementations, from their IT systems. It is not a ban on private Americans or private companies using DeepSeek on their own devices.
When was the order issued, and when does it take effect?
The order is dated July 24, 2026. It was filed for public inspection on October 19, 2026, and is scheduled for formal publication in the Federal Register on October 20, 2026, as document 2026-21447 under docket DHS-2026-0628. Removal is described as immediate, and the 60-day window for agencies to request an exception in writing had already passed before the public filing appeared.
Why does the ban cover apps that are not made by DeepSeek?
Because DeepSeek's models are open-weight, many outside developers built products on top of them. DHS wrote the order to cover "third-party products and services developed using DeepSeek implementations," so an agency cannot keep a DeepSeek-based tool simply because it carries another vendor's name. That is why agencies and their suppliers now have to audit which models are inside their software.
Does the order cover the military?
No. The order is explicitly limited to civilian executive branch agencies and does not cover the Department of Defense. Separate military rules already exist — the US Navy, for example, directed its personnel not to use DeepSeek back in 2025 over security concerns.
Should I delete DeepSeek from my personal phone?
That is your decision — the order does not require it, and this article is not telling you to. The sensible middle ground is to be deliberate about what you share: avoid pasting ID scans, financial documents, work files or private photos into any cloud AI app whose data practices you have not checked, and keep that material in local storage like Pro Calculator Secure Vault instead, with its intruder selfie, incognito private browser and 8 smart calculators (standard, floating, BMI, age, currency, tip, GST & VAT and loan EMI) alongside the vault.
The Bottom Line
Monday's filing is bigger than one Chinese chatbot. It is the US government formally treating an AI model the way it once treated suspect telecom hardware: as supply chain, subject to exclusion, removal and lineage audits across every system it touches. DeepSeek can keep winning benchmarks — its newest model just did — and it will not matter for federal work unless the trust question changes. For agencies and contractors, the work of finding DeepSeek inside their software starts now. For everyone else, the story is a clear reminder to read every AI app's data deal before handing it your life in prompt form. Pro Calculator Secure Vault is built by ZAS — you can see more of our apps on our developer page on Google Play.
No comments:
Post a Comment