Deepseek, a Chinese app that caused $ 1 trillion of US market meltdown this week, preserves rapidly growing U.S. user data.
The artificial intelligence chatbot broke through the chart of Apple's App Store and Google Play Store on Tuesday. According to Appmagic, DeepSeek has been downloaded more than 2 million times since his debut on January 15, and most of the past three days have appeared.
Rival chatbots, including Chatgpt, collect a huge amount of user data, but it is important for Americans to use Chinese-based servers by DeepSeek (created by Math Geek Hedge-Fund investor Liang Wenfeng). Experts said that it was a difference and an obvious privacy risk.
“What makes this context stand out is that Deepseek is a Chinese company -based Chinese company,” said Angela Zhang, a law professor at the University of Southern California, focusing on Chinese technical regulations. I mentioned it.
“This raises the question of whether collecting data such as IP addresses and key stroke patterns can lead to a threat to national security,” Zhang added.
DeepSeek's Terms of Use User data has been preserved “a safe server in the People's Republic of China”. The company also states that it automatically collects personal information such as “device model, operating system, keystroke pattern or rhythm, IP address, and system language”.
All of the company based in China is subject to the Cyber Security Law of the CCP and is required to share data with the government in response to the request.
Chinese chatbots also show signs of censorship and prejudice. Including Chinese leaders Xi Koni, the massacre of the 1989 Tiananmon Square, whether Taiwan is the country, and the prompt to answer the prompt of whether China has committed human rights abuse against the new Jiang Uyghur. 。
In some cases, Deepseek begins to generate a response -stops in the middle of the sentence, and such a prompt is only to write “over there”. [its] Current range. “
Chinese data storage was an important concern that US lawmakers spurred to pursue Tactoku's ban. This was enforced this month after the Chinese parent ordinance could not sell the shares before the deadline on January 19. President Trump has issued a Presidential Ordinance to delay the ban.
This week, the House of Representatives Superat Selection Committee on China, which led Tactoku's accusation, Retweet images depicting DeepSeek As “Troy wood horse”.
“We should work to quickly export more powerful export management in Deepseek's AI infrastructure in DeepSeek,” said the chairman of the committee chairman, John Mulenal (R-Mich).
“The United States allows CCP models such as Deepseek to be at risk for national security, and cannot promote AI's ambitions by utilizing our technology.” Mulenal stated in a statement.
Josh Kushner, the main investor of Openai, has been built using US technology and torn colleagues who publicized DeepSeek.
“The” Pro -America “engineer uses chips that are likely to violate export control and openly support Chinese models trained from the major models of the US Frontier model in the United States. Returns US customer data to China. ” KUSHNER wrote it in X on Monday。
One day ago, Elon Musk tweeted that DeepSeek could access a considerable amount of advanced NVIDIA chips. SCALEAI's Alexandr Wang told CNBC that the company has 50,000 advanced chips that are not publicly recognized for export management.
Wedbush analyst Dan Ives stated that The model of the DeepSeek is “NVIDIA NEXT GENERATION Hardware, and those built for $ 6 million are probably fictional stories.”
Deepseek did not immediately return the comments on the post. The AI company does not seem to have a dedicated public relations team yet.
According to Adrianus Warmenhoven, NordVPN cyber security expert, “Deepseek's privacy policy that can be found in English is clear.
“Users need to consider whether the interaction or uploaded data may be inadvertently contributed to the machine learning process, which may lead to the development of tools that may be malicious or malicious. Warmenhoven added.
Openai's Sam Altman acknowledged that Deepseek was an “impressive model”, but swore that his company would be “clear” in the future release.
In comparison, Openai's GPT-4 cost more than $ 100 million, but Anthropic's DARIO Amodei suggests that last year's currently developing models may cost more than $ 1 billion. did.
Reliable has become a tank due to the US market, and an investor is working on the possibility that advanced hardware built on the AI model has not been considered before. The value has been erased over $ 1 trillion.
He also questioned many people whether the United States had lost AI's advantage of China.
Mark and Leisen, a billionaire technical investor, said that the rise of Deep Shek was the “moment of AI's SPUTNIK.” It mentioned the release of the Soviet Earth Earth satellite in 1957, surprising the United States, and caused a space race between the two superpower.
