Trump Admin Bans DeepSeek From Federal Agencies

The Department of Homeland Security just banned the use of DeepSeek’s technology across much of the federal government. In a notice posted on Monday, the DHS ordered federal civilian agencies like the Department of State to remove any products from the Chinese AI lab DeepSeek, and any third-party services that rely on DeepSeek AI models. “DeepSeek’s account generation process collects personally identifiable information (PII) and may also track and digitally fingerprint users,” Homeland Security Secretary Markwayne Mullin said in an accompanying note from July that was made public only on Monday. The administration fears that this sensitive data collected by DeepSeek could then be transferred to the Chinese government. The models themselves can also include malware, Mullin warned. DeepSeek use has already been restricted across several parts of the government. The Chinese AI lab caused mayhem in the American AI industry early last year with the release of its R1 model, which boasted high capabilities at a fraction of the cost of its American counterparts. While the stock market was busy experiencing a trillion-dollar wipeout over the release, the U.S. government deemed DeepSeek a “profound threat” to national security, and both the Navy and the Department of Commerce were quick to instruct personnel not to use DeepSeek’s models. American AI models still lead Chinese AI development, but that gap is now shrinking. The first signal of this Chinese catch-up was delivered by none other than DeepSeek’s R1 last year, and fellow Chinese AI lab Moonshot reignited these fears in July with Kimi K3. As the competition heats up, so have debates over the potential national security implications of losing the so-called AI race. Though the two superpowers have recently promised increased AI cooperation through a potential framework called the US-China AI Dialogue, China hawks in Washington continue to warn that Chinese AI advancements can give Beijing and the Chinese military the upper hand. More recently, DeepSeek has also been accused of distillation, a process in which a more advanced “teacher” model’s output is used directly to train another, less-advanced “student” model. The process is by no means new and is actually fairly common in the AI industry, but American AI giants like Anthropic are accusing Chinese rivals like DeepSeek of “illicitly” violating their platform rules to improve their own models via this process. The American government has also deemed the alleged distillation attempts “adversarial,” equating it to stealing training data or cheating. The National Security Agency, Cybersecurity and Infrastructure Security Agency, and the FBI accused a host of Chinese AI companies of engaging in industrial-scale distillation campaigns in a joint letter published last month, which also specifically named DeepSeek. Treasury Secretary Scott Bessent even went so far as to threaten sanctions over the practice, which he deemed “IP theft.” Beijing is also reportedly concerned about the national security implications of AI, and wants to keep DeepSeek on a tighter leash. Earlier this month, Beijing authorities proposed measures that could put export controls on the overseas deployment of some Chinese AI models.