OpenAI Proposes Banning DeepSeek Language Models in the US

As is often the case with advanced technologies for China, DeepSeek’s developments quickly found detractors beyond its borders. The rival American OpenAI has proposed banning the use of language models in the US not only by DeepSeek, but also by all companies supported by China or other unfriendly countries.

Image source: OpenAI

This proposal, as TechCrunch notes, was sent by OpenAI management to the US authorities as part of the discussion of the so-called “AI Action Plan”, which is intended to become an American national initiative in this area. OpenAI claims that DeepSeek language models are not safe for the US due to their compliance with Chinese authorities’ requirements for handling user data. At the same time, there is a call to ban all language models from countries unfriendly to the US, since this will preserve data privacy and eliminate security risks, including theft of intellectual property.

It is difficult to say what part of DeepSeek’s developments may actually be banned in the United States. The Chinese developer adheres to the strategy of using open-source models, but this does not mean that third parties using them automatically risk the privacy of their data. Corporate users of such models have every opportunity to ensure the privacy of information. In fact, the only conditional risk is represented by calls to the DeepSeek API, which processes requests on the side of the company’s own infrastructure.

OpenAI has previously accused DeepSeek of using its language models to distill Chinese. Now, the traditional trump card of expressing concerns about information security, which will inevitably be tied to US national interests, is being played.