Openai calls Deepseek “controlled by the state”

In a new policy proposalOpenai describes the Ai Chinese laboratory In depth As a “state -subsidized” and “controlled by the State” and recommends that the American government plans to ban models of the holding and the similar China Republic (PRC).

The proposal, a submission For the Trump administration “Ai Action Plan“Initiative,” says Deepseek models, including its Model R1 “reasoning”are not safe because Deepseek faces the requirements under Chinese law to comply with user data requests. Prohibit the use of “RPC” models in all countries considered “level 1” under the Biden administration export rules would prevent confidentiality and “security risks”, says Openai, including the “risk of IP flight”.

It is not clear if OpenAi’s references to the “models” are supposed to refer to the Deepseek API, the open models of the laboratory, or both. The open models of Deepseek do not contain mechanisms that would allow the Chinese government to siphon user data; companies of which Microsoft,, PerplexityAnd Amazon Hosted them on their infrastructure.

OPENAI A Presented of accused Deepseek, who took importance earlier this year, knowledge of the “distillation” of OpenAi models against its service conditions. But the new allegations of Openai – that Deepseek is supported by the RPC and under his command – are an escalation of the company’s campaign against the Chinese laboratory.

There is no clear link between the Chinese government and Deepseek, a spin-off of a quantitative healing fund called High-Flyer. However, the RPC has been interested in Deepseek in recent months. Several weeks ago, the founder of Deepseek Liang Wenfeng met the Chinese leader Xi Jinping.