Why It Matters
The Department of Homeland Security has ordered DeepSeek technology removed from most civilian federal government systems. The restriction also covers software built with DeepSeek artificial-intelligence models.
What Happened
DHS made the order public Monday, with the ban taking effect at the end of July. It applies to civilian agencies, including the departments of the Treasury, State and Justice.
Agencies within the Department of Defense are excluded from the order. DeepSeek is a Chinese firm that developed an open-source AI model, which was first released publicly in 2025.
DeepSeek released its latest version last week. Its models are incorporated into products or services used by U.S. technology companies including Snowflake, Vercel and OpenRouter.
Michael Kratsios wrote that American AI companies had faced “malicious exploitation.” Anthropic and OpenAI have claimed that some Chinese AI firms copied their models.
By the Numbers
2025 — year DeepSeek initially released its model publicly
End of July — effective date of the government ban
$6m (£4.5m) — amount DeepSeek said it spent training its debut model
$100m — approximate cost associated with developing OpenAI’s GPT-4
Last week — release timing of DeepSeek’s most recent version
Zoom Out
DeepSeek’s reported training cost has drawn comparison with the roughly $100 million associated with OpenAI’s GPT-4 development. The company’s servers have also gone down during periods of high demand.
Donald Trump views China as the key U.S. rival in AI development. Closed-model alternatives named in the broader market include Google’s Gemini, Anthropic’s Claude and OpenAI’s Codex.
What’s Next
Civilian agencies covered by the order must remove DeepSeek technology and tools built with its models. The restriction does not extend to agencies within the Department of Defense.