DeepSeek was set to be added to US Entity List for supporting China’s military and intelligence operations, report claims — White House holds off to avoid escalating tensions with China

White House
(Image credit: Getty / Kevin Carter)
2

AI outfit DeepSeek, along with over a hundred other Chinese companies, was slated for addition to the Department of Commerce’s Entity List last year, according to a new report. An interagency committee purportedly recommended the addition of the Chinese AI startup after a senior U.S. State Department official said that the firm supported Chinese military and intelligence operations. Alongside DeepSeek, Chinese memory maker CXMT was also slated for the blacklist that prevents American institutions from doing business with them. However, Reuters reports that the White House has held off updating its blacklist to avoid escalating trade tensions with Beijing, especially in the lead-up to President Donald Trump’s three-day state visit to China.

Aside from being accused of helping the Chinese military, Anthropic also claims that DeepSeek and two other frontier Chinese models distilled Claude to improve the capabilities of its models, making 16 million exchanges with 24,000 fraudulent accounts. The American AI company said in its statement on X, “Distillation can be legitimate: AI labs use it to create smaller, cheaper models for their customers. But foreign labs that illicitly distill American models can remove safeguards, feeding model capabilities into their own military, intelligence, and surveillance systems.” There have also been reports that the AI firm has been using shell companies to acquire banned Nvidia chips.

Latest Videos FromTom's Hardware
Jowi Morales
Contributing Writer

Jowi Morales is a tech enthusiast with years of experience working in the industry. He’s been writing with several tech publications since 2021, where he’s been interested in tech hardware and consumer electronics.

  • hotaru251
    Anthropic also claims that DeepSeek and two other frontier Chinese models distilled Claude to improve the capabilities of its models, making 16 million exchanges with 24,000 fraudulent accounts. The American AI company said in its statement on X, “Distillation can be legitimate: AI labs use it to create smaller, cheaper models for their customers. But foreign labs that illicitly distill American models can remove safeguards, feeding model capabilities into their own military, intelligence, and surveillance systems.”

    funny way to whine about being used to train it w/o permission given the entire business model was trained w/o asking others.... as well as calling em out for working with military, intelligence, and surveillance systems.....which is done by more than just China ..example being USA using it for military and surveillance done via palantir
    Reply
  • ClemCa
    hotaru251 said:
    funny way to whine about being used to train it w/o permission given the entire business model was trained w/o asking others.... as well as calling em out for working with military, intelligence, and surveillance systems.....which is done by more than just China ..example being USA using it for military and surveillance done via palantir
    The US applies a strong "rules should apply to others" kind of policy, where anything they consider unfair shouldn't be done by others, but if it's the US it's fair game.
    Reply