DHS Bans DeepSeek from US Civilian Agencies Over Data Security Risks

Key points

• DHS ordered the removal of DeepSeek from civilian agency IT systems, including the Treasury, State, and Justice Departments.
• The ban prohibits contracting DeepSeek products and using third-party services built with its models.
• DHS stated that DeepSeek models present 'multiple threat vectors' and could lead to sensitive data transfer to China.
• The Department of Defense is excluded from this specific ban, though it has previously restricted DeepSeek use.
• DeepSeek’s debut model cost approximately $6 million to train, compared to $100 million for OpenAI’s GPT-4.
• The ban follows similar restrictions by the Commerce Department and the US Navy in the previous year.

Background

This action occurs amid escalating US-China tech tensions. In September 2026, China enacted new regulations allowing authorities to ban citizens from leaving the country for up to three years if they violate technology export controls or pose national security risks. These measures aim to retain tech talent and capital, reflecting Beijing’s strategy to support its innovation goals and compete with the US in AI and semiconductors. The current ban on DeepSeek is part of a broader US effort to mitigate perceived security risks from Chinese AI firms, which have been accused of 'distillation attacks' on American models.

How outlets are covering it

BBC and Politico agree on the core facts of the DHS ban but emphasize different aspects. BBC highlights the economic disparity, noting DeepSeek’s low development costs and its popularity among US startups and tech companies like Snowflake and Vercel. It also mentions criticism from US firms like Anthropic and OpenAI regarding model copying. Politico focuses on the security rationale, quoting DHS’s claim that DeepSeek models 'present multiple threat vectors' and could result in sensitive data transfer to China. Politico also notes that the ban follows similar actions by the Commerce Department and the US Navy, framing it as part of a broader federal effort to limit Chinese AI usage. Both sources acknowledge that DeepSeek’s open-source nature and low cost make it attractive, but they differ in emphasis: BBC on market impact and competition, Politico on security and regulatory consistency.

Why it matters

The ban underscores the US government’s prioritization of national security over cost efficiency in AI adoption. It signals a shift toward restricting Chinese AI technologies in critical government functions, potentially impacting the development and deployment of AI in civilian agencies. The move also highlights the tension between the economic benefits of open-source AI and the security risks associated with foreign-developed models. As US firms like OpenAI and Anthropic prepare for public listings, the ban may influence market dynamics and investor confidence in AI companies with Chinese ties.

What to watch

The DHS has not detailed the specific security concerns beyond the general threat of data transfer. It is unclear whether the ban will extend to other Chinese AI firms or if similar restrictions will be imposed on private sector use of DeepSeek. The exclusion of the Department of Defense suggests that military agencies may have different protocols for AI usage. Future developments may include further regulatory actions against Chinese AI firms, potential legal challenges from companies using DeepSeek, and increased scrutiny of AI data security in government contracts.