The American House Council Committee says Dibsic is a Chinese artificial intelligence “a deep threat” to national security

The Parliament Committee from both parties recommended Wednesday Putting restrictions on exporting artificial intelligence models To China after concluding that Deepseek trained its low -cost models using data from Openai’s Chatgpt. He also suggested imposing a ban on federal agencies that buy artificial intelligence models from China, which does not seem to be something that will happen anyway.

The Chinese House selection committee concluded that Dempsic poses a “deep threat” to American national security by collecting user data from Americans and sending it to China. Legionships claim that the founder of Dembesic, Liang Winfing, controls Depsik as well as a hedge box called high -style, with relationships with the state -related research laboratories.

The new report comes a day after the Trump administration was placed Restrictions on the export of chips including Nvidia’s H20 For China, there is a significant big blow to the company that made billions to sell sealed chips, but it is still able to address artificial intelligence in the Chinese market. It was previously believed that the administration would continue to allow the free flow of the lower finish chips in the country after the CEO Jensen Huang Make the pilgrimage to Mar LagoBut it seems that President Trump has committed a rug as licenses will be required now. Critics suggest that this step is not studied, as NVIDIA managed to make billions of dollars while selling the treatments of Chinese companies that were less than available to the Western world. By reducing access to these low chips, China will motivate to build its own chips and faster, and NVIDIA will lose money that can keep America in the future. It seems that the Trump administration does not agree and wants to investigate whether NVIDIA Provide more technology For Chinese clients from what he revealed.

Openai was the largest supporter of Microsoft Quickly crying wrong After the launch of the R1, claiming that it was trained using the distillation technology that includes studying another model output, including thinking steps, to quickly reach speed. The paradox should not be lost on anyone that a notorious company due to taking content from all over the web without permission has now attacked a competitor to doing the same. There is no small violin enough for Openai.

What’s more, the report of the new parliament committee, which seeks to curb Deepseek, was not affected by anything other than Openai, Bloomberg reports:

In its report, which referred to a certificate from Openai, the committee claimed that Deepseek is likely to use “illegal” distillation techniques, as the AI ​​model uses other outputs for training purposes to develop similar capabilities.

Openai told the committee that Deepseek staff “defrauded the handrails” to extract thinking outputs and accelerate development at a lower cost and claimed that the Chinese startup has used its models “to classify model responses and liquidate and transfer training data.” “It is possible that” Deepseek also uses the leading AI models to create high -quality artificial data. “

R1 is an open source model, which means that it can be hosted locally on American servers and code can be studied and modified. American companies including microsoft themselves as well as definition and confusion are The model is already hosting On their cloud platforms. It is true that consumers visit Chat.deepseek.com It will interfere with a China -based counterpart of the Chatbot model, but the model itself is not hateful of its nature.

If the previous examples of America restrict high -tech exports to China are any indicator, the increase in reducing artificial intelligence from entering the country may not do much to slow down the country and may actually help them. Current restrictions have forced the export of advanced chips, such as Deepseek, to create highly efficient models on low processors. When Huawei was cut off from working with semiconductor manufacturers, it only accelerated the plans by that company to make its own slides.

The United States government fears that China will employ Amnesty International to enhance its geopolitical goals, and this is a fair concern. The models developers control the types of content they will discuss, and the Chinese Communist Party can use the training step to influence the viewpoint of the model in its favor. It seems unlikely to be stopped developing artificial intelligence in China at this stage.

The Trump administration also continues its work to make the American Tijocs independent of its Chinese owner, but reports indicate that the government there will not discuss this possibility until the definitions are raised.