Is DeepSeek Safe to Use? Experts: “Absolutely Not”

On Monday, a new generative AI app called DeepSeek became available for download in the United States. The Chinese-created app quickly became the most downloaded free app in the US. Its quick spike in popularity has promoted concerns about data security as well as questions about the US’s dominance in AI.
Security Concerns About DeepSeek AI
Users have reported that DeepSeek is trained to avoid politically sensitive questions that are often censored in China, raising questions about its accuracy and its political stances. Many other Chinese apps, such as Baidu’s Ernie or Doubao by ByteDance, exhibit this same sort of censorship, which is standard practice in China. For instance, DeepSeek users have said that the app refused to give details about the Tiananmen Square protests and massacre, a taboo subject in China. 
DeepSeek also claims that Taiwan is an “unalienable” part of China (the People’s Republic of China [PRC] considers Taiwan a “breakaway” territory, even though it’s never been governed by China).
US officials have also expressed concerns about the generative AI’s data security and privacy issues. According to DeepSeek’s privacy policy, the app collects technical information such as keystroke patterns and IP address as well as diagnostic information like crash reports. It also tracks “the features you use and the actions you take” and stores that data on servers located in the PRC.
Chinese laws give the government authority to access data (including user data) from companies based in China, which includes DeepSeek. This means that the Chinese government could leverage personal data from US users for its own ends, whether that’s training AI models or creating targeted disinformation campaigns.
Why DeepSeek Made US Tech Stocks Plummet
DeepSeek’s sudden rise caused US tech stocks to take a tumble. Most notable was NVIDIA, which dipped 17 percent, roughly equivalent to losing $600 billion off its market value. It set a record for the largest single-day drop in a US stock.
This happened because DeepSeek claimed that it developed its AI model for only $6 million, a small portion of the hundreds of millions of dollars spent by US AI companies such as OpenAI. Experts theorize that this was possible because DeepSeek paired more advanced computing chips with smaller, less expensive ones.
Since NVIDIA is one of the largest manufacturers of those more advanced chips, investors are now questioning whether or not demand for the more expensive chips will continue to climb or not. Many are also wondering whether US tech companies actually need to invest billions of dollars into creating and training AI models and infrastructure, when a Chinese company was allegedly able to create a similar model for a fraction of the cost.
The post Is DeepSeek Safe to Use? Experts: “Absolutely Not” appeared first on eWEEK.