By Katrina Manson and Jordan Robertson
January 30, 2025 at 2:28 PM EST
Updated on January 31, 2025 at 1:34 PM EST
US Defense Department employees connected their work computers to Chinese servers to access DeepSeek’s new AI chatbot for at least two days before the Pentagon moved to shut off access, according to a defense official familiar with the matter.
The Defense Information Systems Agency, which is responsible for the Pentagon’s IT networks, moved to block access to the Chinese startup’s website late Tuesday, the official and another person familiar with the matter said. Both asked not to be named because the information isn’t public.
The move came after defense officials raised concerns that Pentagon workers were using the tool, the person said. DeepSeek’s privacy policy states that it stores user data on servers in China and that it governs that information under Chinese law.
On Wednesday, some Pentagon work screens showed a sign saying “Website Blocked,” citing operational reasons, but others could still access DeepSeek, according to the defense official and correspondence reviewed by Bloomberg News.
A spokesperson for the Defense Information Systems Agency didn’t respond to a request for comment.
Chelsea Dietlin, a Department of Defense spokesperson, said the Pentagon was aware of reporting on DeepSeek usage. As a matter of practice and operational security, the department doesn’t comment on the status of its networks and systems, Dietlin added.
DeepSeek and the AI model that it says it developed for less than $6 million upended markets around the world earlier this week, reigniting questions about the hundreds of billions of dollars that major US tech companies are spending on AI infrastructure. Tech executives including Marc Andreessen praised DeepSeek, and the model soared to the top of the Apple Store downloads. At the same time, security concerns mounted over the chatbot, including the potential misuse of users’ information being stored at Chinese data centers.
The Pentagon’s IT experts are still determining the extent to which employees directly used DeepSeek’s system through a web browser, the official said.
US military personnel started downloading an earlier release of DeepSeek code on their workstations in the fall of 2024, according to the person familiar with the matter. At the time, the downloads didn’t raise concern with Defense Department security teams as the connection to China wasn’t clear to them, the person added.
The explosion of interest in the latest DeepSeek release has triggered efforts in some of the military services to find and remove code from China-origin chatbots on employees’ individual machines, according to the person.
However, thousands of Defense Department personnel are using DeepSeek through Ask Sage, an authorized software platform, that doesn’t directly connect them to Chinese servers, according to Nicolas Chaillan, founder and chief executive officer of the platform.
Military services are addressing DeepSeek use by employees in different ways. The Navy on Friday prohibited any usage of DeepSeek due to potential security and ethical concerns associated with the model’s origin and usage, according to CNBC.
Asked about the report, Navy spokesperson Lieutenant Commander Lauren Chatmas said the Navy already has guidance against using open source AI systems for official work. Recent internal Navy correspondence mentioned DeepSeek in connection with that guidance, she added.
The Department of the Air Force has no specific guidance on DeepSeek but already prohibits the use of sensitive public information in commercial generative AI systems without proper approvals, according to Laura McAndrews, an Air Force spokesperson. The Army issued guidance in June 2024 warning of “unique challenges in terms of data privacy, security and control over the generated content,” according to a public memo, which urged individual commands to develop appropriate governance processes while discouraging banning generative AI tools outright.
Some service members are now discussing the possibility of issuing new policies to explicitly ban Chinese generative AI models, according to correspondence reviewed by Bloomberg News.
https://www.bloomberg.com/news/artic…s-before-block
Deepseek mengguncang dunia