DeepSeek reportedly collects IP, keystroke patterns, device info, and stores it in China
Chinese AI Startup DeepSeek Limits Registrations After Cyberattack Amid Surging Popularity
Chinese AI startup DeepSeek announced on Monday that it would temporarily limit registrations following a cyberattack, just as its AI assistant experienced a sudden surge in popularity.
Earlier in the day, DeepSeek’s website suffered outages after its AI assistant became the top-rated free app on Apple’s U.S. App Store. The company has since resolved issues with its API and login problems, according to its status page. Monday’s disruptions marked the longest service interruption in 90 days, coinciding with DeepSeek’s explosive rise in popularity.
Last week, DeepSeek launched a free AI assistant that it claims operates with significantly lower data and costs compared to leading AI models, potentially signaling a shift in the investment required to develop cutting-edge AI.
The application is powered by the DeepSeek-V3 model, which its creators claim outperforms all open-source models and competes with the most advanced closed-source AI models globally. Since its January 10 release, the app has gained massive traction among U.S. users, according to Sensor Tower, an app data research firm.
Challenging U.S. AI Dominance
DeepSeek’s rapid rise has shaken Silicon Valley, challenging long-standing beliefs about U.S. supremacy in AI and the effectiveness of Washington’s export controls on China’s AI and semiconductor sectors.
On Monday, technology stocks took a hit, with shares of Nvidia (NVDA.O) and Oracle (ORCL.N) plummeting. AI models like ChatGPT and DeepSeek require advanced chips for training, and the Biden administration has progressively tightened restrictions on the export of AI chips to China since 2021.
However, DeepSeek researchers stated in a paper last month that the DeepSeek-V3 model was trained using Nvidia’s H800 chips—a less powerful version of Nvidia’s high-end AI processors—spending less than $6 million in the process. This claim has been disputed, but if true, it suggests that Chinese firms may be finding ways to develop competitive AI models despite U.S. chip restrictions.
A Rising Force in China’s AI Boom
DeepSeek remains shrouded in mystery, operating as a small Hangzhou-based startup founded in 2023—the same year Chinese tech giant Baidu launched China’s first large-language AI model. Since then, dozens of Chinese companies have entered the AI race, but DeepSeek is the first to be widely recognized by U.S. tech experts as potentially matching or surpassing the performance of cutting-edge American AI models.