US bans DeepSeek from civilian agencies over AI security — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Department of Homeland Security ordered all civilian government agencies to remove DeepSeek’s AI technology from their systems, effective end of July, with the notice made public on 23 September
- DeepSeek operates an open-source AI model that has gained rapid popularity, with its chatbot among the most downloaded apps on the Apple App Store and its models used by US tech firms like Snowflake and Vercel
- White House tech lead Michael Kratsios stated in an internal memo this year that Chinese AI firms have engaged in 'malicious exploitation' of American AI technology, naming China as a primary threat
- Anthropic and OpenAI have accused Chinese firms, including DeepSeek, of copying their AI models, as both US companies prepare for public stock listings
- DeepSeek claimed it cost approximately $6 million to train its debut model, a fraction of the roughly $100 million OpenAI spent on GPT-4, contributing to cost advantages that disrupted the AI market
- The ban applies to civilian agencies such as the Department of the Treasury, State, and Justice, but explicitly excludes Department of Defense agencies from the restriction
Why it matters: US tech startups and firms relying on low-cost open-source AI now face uncertainty in access, while the exclusion of the Pentagon suggests a strategic split in how AI threats are assessed across government. The $94 million cost gap between DeepSeek and GPT-4 training underscores a material shift in AI development economics.
Ask SkimNews
