US bans DeepSeek AI from government use as tech tension with China continues
BBC News ● Covered by 2 sources
The US has banned DeepSeek AI from most civilian government systems. It’s a big hit to a cheap, open Chinese model that’s already popular with startups and tech firms.
Based on reporting by BBC News — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The US has ordered DeepSeek’s AI to be stripped out of most civilian government systems, escalating Washington’s pushback against Chinese AI. The Department of Homeland Security put the ban in public on Monday, but said it had already taken effect at the end of July.
The order reaches all use of DeepSeek inside civilian agencies, and also covers any technology built with its models. That includes the Department of the Treasury, the Department of State and the Department of Justice. Military agencies are left out.
DHS did not spell out the security worries behind the move. But the White House has been sharply critical of AI firms from China this year, and Trump administration officials have framed China as the main US rival in the AI race. In an internal memo earlier this year, White House tech lead Michael Kratsios said “malicious exploitation” of American AI firms had been happening, mainly by rival firms in China.
DeepSeek has become a serious nuisance for that industry. Its models are open source, cheaper to use and widely adopted by startups. The company’s chatbot app has also climbed into the most-downloaded ranks in Apple’s App Store, while US companies including Snowflake, Vercel and OpenRouter have built DeepSeek into their products.
And the friction isn’t just about government policy. Anthropic and OpenAI have both said Chinese firms, including DeepSeek, have effectively copied their models. DeepSeek’s own numbers helped fuel the tension: last year it said training its debut model cost about $6m, far below the roughly $100m OpenAI has said was spent on GPT-4. The latest DeepSeek model arrived last week, and users online have been praising it for strong coding performance.
My take — AI-written commentary, not fact-checked reporting
The US is doing what governments always do when a cheap rival gets too popular: calling it a security issue. Maybe there are real risks, but this also looks like a market-protection reflex dressed up in policy language. Open models are the pressure point, and Washington clearly hates how much leverage they give away.
Read more about this at: BBC News
