The Deepseek AI application is seen on a mobile phone in this photo illustration taken in Warsaw, Poland, January 27, 2025.
JAAAP Arriens | Nurphoto | Getty images
The startup of Chinese artificial intelligence Deepseek could face borders of the American government because it upsets the ecosystem of the American AI, although the application of restrictions on open source technology can be a challenge, have declared experts .
The sudden rise in Deepseek has questioned the effectiveness of Washington’s efforts to limit China access to high -end technology compared to national security problems.
For Deepseek, which was based strongly on the open-source code, the additional export restrictions that the United States government could impose are limited, said Lawrence Ward, partner of the American law firm Dorsey & Whitney with specialization in law national security.
Regarding the potential use by the company of certain Nvidia chips, it could face civil and criminal sanctions, but the application of these sanctions can be “difficult, if not impossible,” said Ward.
US legislators on Monday called for actions to slow down The Chinese technological startup, with some calling Deepseek “a serious threat”.
There are no easy solutions to restrict the use of an open-source model, in particular that which is widely tested and used by organizations and individuals, said Paul Triolo, associated with Albright Stone Group.
An option would be that the Commerce Department develops rules that require technology giants such as Apple and Google to eliminate the Deepseek application, restricting its downloads to the American market, said Triolo, adding that it would be however Difficult to remove the application from other platforms such as GitHub.
Deepseek in December published a model of large free and open-source languageWhich, according to him, was built in just two months and a fraction of the cost borne by other major actors. Last week, the company published a model of reasoning that would have surpassed The latest Openai offers in many third -party tests.
Deepseek, whose chatbot Ai has become the Most free application downloaded On the Apple App Store in the United States, said that it had created the model despite the scanning orders imposed in 2022 by the United States on China access to The most advanced Nvidia fleas.
The Hangzhou -based startup said that she used the Less advanced nvidia h800 chips To build the model, which was available until the United States The government expressed the ban in October 2023.
The situation is somewhat unprecedented. We are clearly in unexplored waters on several levels.
Paul Triolo
Partner, Albright Stone Group
“If Deepseek had equipment that it shouldn’t have, this could certainly lead to a more in -depth investigation,” said Daniel Newman, CEO of Futurum Group, who expects the US government to examine Deepseek equipment which could have raped the export borders.
President Donald Trump said on Monday that the progress of the application of Chinese AI Deepseek “should be an alarm clock” For America’s technological companies, while renting the low -cost model as “positive development” for AI overall. “Instead of spending billions and billions, you will spend less and you will hope, the same solution,” said Trump.
Deepseek’s profitability and efficiency claimed from its model have sent American technological actions to a tailspin, because investors have questioned the amount of money that large technological companies have invested in AI models and centers data. American companies could now take advantage of public codes to refine their own models and possibly reduce calculation costs.
Representative John Moolenaar, R-Mich., Chairman of the China Selective Committee, said on Monday that the US government should strive to place stronger export controls on essential technologies for IA infrastructure Deepseek.
“The situation is somewhat unprecedented, and it is unlikely that anyone in Washington has a clear idea of what to do on this subject … We are clearly in unexplored waters on several levels,” said Triolo.
The company said on Monday that it New registrations temporarily limited Users with continental China telephone numbers following a large -scale malicious cyber attack that caused breakdowns on its website.
On January 20, the day when the company’s R1 reasoning model was launched to the public, the founder of Deepseek Liang Wenfeng attended a closed -door symposium for the businessman and the expertsChaired by Chinese Prime Minister Li Qiang, according to the Xinhua state media.
Dylan Butts of CNBC, Hui Jie Lim contributed to this report.
