Depositphotos
The Pentagon has blocked internal access to DeepSeek’s AI models after discovering that employees in some departments had been connecting to Chinese servers for at least two days, reports Bloomberg.
According to DeepSeek’s terms of service, user data is stored on Chinese servers, and the country’s legislation is known to require local companies to cooperate with intelligence agencies. On Wednesday, some Pentagon desktop screens displayed the sign «website blocked» when trying to launch DeepSeek.
The publication also clarifies that «thousands of Defense Department employees» used DeepSeek through Ask Sage — an authorized software platform that does not connect them directly to Chinese servers. At the same time, one of Bloomberg’s sources says that the US military first tried to «download the previous release of DeepSeek to their workstations in the fall of 2024».
Previous use of the Chinese AI model by employees banned by the US Navywhile in Italy the blockade was extended to the whole country.
This month, DeepSeek disrupted the artificial intelligence industry with the release of model of reasoning R1which allegedly outperformed such competitors as o1 by OpenAI in some key tests; and, given that the Chinese startup offered free access, its app quickly overtook ChatGPT in the App Store ranking and PlayStore. A research article by DeepSeek, which claimed that its model was trained on a cluster of 2,000 Nvidia chips and cost only $6 million, while competitors needed many times more, added to the problem caused the collapse of shares leading developers of processors for artificial intelligence and large technology companies.
After the hype died down, it turned out that DeepSeek not only collects user information and provides misleading answers about some events in China, but also actually kept a large database in the public domain. OpenAI has also recently announced that it has evidence that Chinese AI models trained on ChatGPT data.