US Bans DeepSeek AI From Government Systems — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- DHS ordered DeepSeek AI removed from all civilian government agencies, with the ban taking effect at the end of July and disclosed publicly on Monday; it covers the Treasury, State, and Justice Departments but excludes Department of Defense agencies
- White House tech lead Michael Kratsios wrote in an internal memo that "malicious exploitation" of American AI firms had been occurring mainly by rival firms in China, though the DHS notice did not detail specific security concerns about DeepSeek
- Anthropic and OpenAI — both poised to list on public stock markets in the coming months — have accused DeepSeek and other Chinese rivals of essentially replicating their AI models
- DeepSeek said it spent about $6m to train its debut model, compared to roughly $100m for OpenAI's GPT-4, a cost gap that sent tech stocks tumbling when revealed last year
- DeepSeek's most recent version, released last week, has been praised online by users who say it outperforms OpenAI's tools on common coding tasks
- DeepSeek's free chatbot app ranks among the most downloaded in the Apple App Store, while its open-source models are incorporated into offerings from Snowflake, Vercel, and OpenRouter
- Trump has framed China as the US's key rival in AI development, calling the technology crucial to America's future
Why it matters: The DHS ban formally excludes DeepSeek from civilian federal workflows even as US startups and tech companies keep building on its models because of a cost gap the ban doesn't touch — DeepSeek trained its debut model for roughly $6m versus about $100m for GPT-4. That price advantage, combined with Anthropic and OpenAI claiming DeepSeek has copied their work, turns the ban into a security marker that leaves the commercial open-source adoption story intact.
Ask SkimNews
