US declines to blacklist DeepSeek despite security reviews

Editorial illustration for: US government holds off blacklisting DeepSeek amid ongoing security reviews

In brief

  • US declined to blacklist DeepSeek despite months of security reviews and congressional legislation.
  • New York, Texas, and Virginia banned DeepSeek from government systems starting January 2025.
  • Congress introduced the No DeepSeek on Government Devices Act in February 2025.
  • DeepSeek faces accusations of storing user data in China and maintaining military links.
  • DeepSeek's V4 model is slated for release in early 2026.

State and Congressional Action Outpace Federal Response

Multiple states including New York, Texas, and Virginia moved to ban DeepSeek from government systems starting in January 2025. In February 2025, lawmakers introduced the No DeepSeek on Government Devices Act, a bill aimed at formalizing restrictions on the firm's software across government networks. These actions signal growing alarm among state officials and Congress over the firm's presence in sensitive government infrastructure.

Yet the federal government has moved differently. When pressed on potential blacklisting actions, US officials have said there is "nothing to announce." The restraint stands out given the scale of concern. More than 100 other entities considered security risks have also avoided being added to the Commerce Department's Entity List, suggesting a broader pattern of caution in formal trade designations.

The DeepSeek Puzzle

DeepSeek's rise created an immediate market shock. The company demonstrated that you didn't necessarily need the most advanced Nvidia chips to build competitive AI, a finding that upended assumptions about the cost and complexity of AI development. That realization triggered a sell-off in AI-related equities in early 2025.

DeepSeek, backed by the Chinese quant hedge fund High-Flyer, burst onto the global AI scene with models that performed remarkably well. But the company faces serious allegations. DeepSeek has been accused of storing user data in China, allegedly attacking US AI models, and maintaining links to the Chinese military. These concerns haven't triggered formal blacklisting—yet.

What Comes Next

Discussions about phased escalation are reportedly ongoing, and the legislative groundwork has already been laid. The federal government appears to be taking a measured approach, even as states and Congress push for tighter restrictions. Meanwhile, DeepSeek's V4 model is slated for release in early 2026, which would represent another leap forward for the firm and potentially reignite the competitive pressures that spooked markets earlier this year.

The gap between state action and federal inaction suggests the government is still weighing how to respond to Chinese AI competition without triggering broader trade escalation or disrupting domestic AI development.