The code for DeepSeek, an impressive artificial intelligence model made in China that stunned Silicon Valley and quickly became one of the most downloaded apps in the U.S., has hidden code that can send user information to the Chinese government, according to security experts.
Researchers at Feroot Security, a Canada-based cyber firm, examined the browser-based version of DeepSeek and said they uncovered encrypted portions of code capable of sending user data to CMPassport.com, the online registry for China Mobile, a state telecom company that’s been delisted from the New York Stock Exchange and is considered a national security concern by U.S. regulators.
The AI app also creates a digital “fingerprint” of users, tracking their activity on all other websites, according to Feroot.
"We see direct links to servers and to companies in China that are under control of the Chinese government. And this is something that we have never seen in the past," Feroot CEO Ivan Tsarynny told ABC News.
The Independent has contacted DeepSeek and China Mobile for comment.
DeepSeek has previously disclosed user data being stored on servers based in China.
Other experts expressed similar dismay over the AI company.
.png?trim=0%2C0%2C0%2C0)
The app “raises all of the TikTok concerns plus you’re talking about information that is highly likely to be of more national security and personal significance than anything people do on TikTok,” former Homeland Security and National Security Agency official Stewart Baker told The Associated Press.
The free, open-source DeepSeek, which debuted last month, sent shockwaves through the tech world, causing more than $1 trillion in losses on the stock market as the company raised questions about the value of more well-capitalized U.S. competitors like OpenAI, which reportedly spent billions more developing similar capabilities.
“DeepSeek’s release of a premium level AI tool, available freely, with a reported (comparatively) miniscule development cost has shaken faith in Silicon Valley and American dominance in the rapidly developing AI market,” Dr. Richard Whittle, an economist from the University of Salford, told The Independent at the time.
In addition to posing a potential challenge to U.S. AI might, DeepSeek also raised geopolitical alarm bells in Washington, with Donald Trump describing it as a “wake-up call” for Americans about the state of Chinese tech power.
Senator Josh Hawley of Missouri introduced a bill this week widely seen as a response to DeepSeek, which aims to “prohibit United States persons from advancing artificial intelligence capabilities within the People’s Republic of China, and for other persons”.
The bill bars the importation of “technology or intellectual property” developed in China, with anyone violating the restrictions facing up to 20 years in prison, and fines of up to $1 million for individuals and up to $100m for businesses.
Users of the app also quickly noted it equivocated or refused to answer certain questions on topics sensitive to Beijing, like Chinese leader Xi Jinping, the subjugation of China’s Uyghur Muslims, or the history of the Tiananmen Square massacre.
The app has already been banned across countries like Italy, as well as within institutions like the U.S. Navy and NASA, and it could soon go the way of TikTok in the U.S., which is facing an impending U.S. ban amid similar concerns over Chinese influence unless Donald Trump can reach a promised compromise forcing a sale of its U.S. operations.
Fears of Chinese AI dominance seem to have inspired tech companies like Google to change their stance on the technology in recent months.
The search giant reportedly altered its AI guidelines recently to allow for use in weapons and security applications.
In a recent company blog post, leaders cited “a global competition taking place for AI leadership within an increasingly complex geopolitical landscape.”
Meanwhile, in December, ChatGPT maker OpenAI announced a partnership with fellow tech company Anduril to develop technology for use in military defense against drones.
OpenAI had previously barred its technology from military use but changed its policies last year to allow some collaborations.
