US bars DeepSeek from federal civilian agencies over China data security fears
The Department of Homeland Security orders federal civilian departments to strip out DeepSeek technology and halt third-party procurements.

The US Department of Homeland Security has ordered all federal civilian agencies to purge Chinese artificial intelligence firm DeepSeek from their IT networks, barring staff and contractors from acquiring or deploying its models.
The directive issued on Monday instructs civilian departments to remove all DeepSeek software, affiliate offerings, and third-party tools built on top of the lab’s architecture. Agencies must strike the technology from procurement catalogues, approved product lists, and vendor contracts. The order does not apply to national security systems or classified networks run by the Department of Defense and intelligence agencies.
Why DHS ordered the ban
Homeland Security officials said DeepSeek’s systems expose US operations to "multiple threat vectors" and risk funnelling sensitive data directly to Beijing. The department pointed to the company’s physical infrastructure, noting that its servers are based in mainland China, where domestic laws give state intelligence and security apparatuses wide authority to access private company data.
The department also flagged privacy and software integrity concerns, warning that DeepSeek collects personally identifiable user information and creates digital fingerprints during account registration. DHS warned that combining that data with public records could yield actionable intelligence on federal personnel and programmes, while adding that the underlying models could harbour malware or undetected vulnerabilities.
Civilian departments have been instructed to tighten data exfiltration controls, monitor network traffic, and log suspicious activity until all DeepSeek technology is uninstalled. Any agency seeking an exemption must submit a written appeal within 60 days, demonstrating that the technology is indispensable to national security or a mission-critical operation.
Growing scrutiny of foreign tech
DeepSeek gained global prominence after releasing open-source models that delivered competitive benchmarks at a fraction of the operating cost charged by American rivals. That cost edge prompted several commercial platforms, including Snowflake, Vercel, and OpenRouter, to integrate its models into their offerings.
However, federal and state scrutiny has mounted steadily. The US Department of Commerce and the US Navy had already barred DeepSeek from official devices, and state governments in New York, Tennessee, and Virginia implemented similar restrictions. Leading US developers, including OpenAI and Anthropic, have also alleged that Chinese competitors distilled their proprietary models to bootstrap their own systems.
Separately, the federal government moved on another tech front on Monday, with the Department of Justice penalising Silicon Valley software firm Cloudera for hiring discrimination. Cloudera agreed to establish a $1.4 million back-pay fund and pay $600,000 in civil penalties after federal investigators found the firm sidelined qualified American applicants across seven job postings to favour temporary visa holders under the Permanent Labor Certification programme. The settlement bars Cloudera from the green card sponsorship scheme for two years.
Key numbers
| What | Figure | Source |
|---|---|---|
| DeepSeek exemption window | 60 days | US Department of Homeland Security |
| Cloudera back-pay fund | $1.4 million | US Department of Justice |
| Cloudera civil penalty | $600,000 | US Department of Justice |
| Cloudera PERM programme ban | 2 years | US Department of Justice and Department of Labor |
In this story
| Name | Role in the story |
|---|---|
| US Department of Homeland Security | Issued the directive ordering civilian agencies to purge DeepSeek systems. |
| DeepSeek | Chinese artificial intelligence lab whose models were banned from US civilian systems. |



