The U.S. State Department has publicly alleged that Chinese artificial intelligence firm DeepSeek is playing a direct role in bolstering Beijing’s military and intelligence operations while sidestepping American export restrictions on advanced semiconductors. According to a senior U.S. official, DeepSeek has moved beyond merely sharing open-source AI models to actively aiding China’s defense apparatus and employing a network of shell companies to obtain banned high-end chips.
DeepSeek, headquartered in Hangzhou, captured global attention earlier this year when it unveiled AI reasoning models—DeepSeek-V3 and DeepSeek-R1—that it claimed matched or outperformed top U.S. industry offerings at a fraction of the cost. While those announcements fueled excitement across the tech world, U.S. authorities have since grown skeptical, concluding that the startup’s rapid ascent may owe more to covert support from the Chinese government and U.S.-origin technology than to breakthrough innovations.
Allegations of Military Collaboration
A senior State Department official, speaking on condition of anonymity, told reporters that DeepSeek “has willingly provided and will likely continue to provide support to China’s military and intelligence operations.” U.S. investigators point to internal procurement records in which DeepSeek appears more than 150 times in contracts linked to the People’s Liberation Army and affiliated defense research institutions. Those records, combined with statements from military procurement officers in Beijing, suggest the firm has supplied bespoke AI services to enhance surveillance, logistics planning, and other defense-related applications.
U.S. lawmakers reinforced these concerns when the House Select Committee on the Chinese Communist Party opened a parallel inquiry into DeepSeek’s corporate structure and technology transfers. A committee report detailed how founder Liang Wenfeng exercises de facto control over the company through a web of holding entities, raising questions about state influence behind DeepSeek’s ostensibly private-sector guise. That report also urged tighter export controls, warning of a “strategic surprise” if Beijing gains unfettered access to cutting-edge AI capabilities.
Circumventing Export Restrictions
Since 2022, U.S. policy has banned the sale of Nvidia’s H100 GPUs to mainland China, citing fears these chips could dramatically boost Beijing’s military-grade AI development. Despite those curbs, U.S. officials allege DeepSeek obtained “large volumes” of Nvidia’s high-performance chips by routing orders through a series of shell companies based in Malaysia, Singapore, and Vietnam. In one documented case, DeepSeek reportedly contracted with a small Singapore-based data center to host American-made GPUs, then accessed them remotely—an approach officials say exploits a loophole that still allows non-Chinese firms to rent U.S. chips in third-country facilities.
While DeepSeek has not been formally blacklisted by the U.S. government, State Department sources indicate that Commerce Department investigators are probing whether the firm violated the International Traffic in Arms Regulations (ITAR) or other export statutes. DeepSeek’s leadership denies any wrongdoing, insisting that all chips were acquired legally and that its models run on a mix of H800 series products—export-approved alternatives to the H100 line. Nvidia has likewise maintained that it adheres strictly to American export laws, asserting that any chips found in Southeast Asia were sold through non-China subsidiaries and without knowledge of their end-use.
Data Privacy and Global Scrutiny
In addition to export control concerns, Washington accuses DeepSeek of funneling user data to Beijing’s vast surveillance network. Under China’s 2017 National Intelligence Law, all domestic tech companies must comply with government requests for information. However, U.S. officials claim DeepSeek voluntarily shares detailed usage statistics, model-training logs, and even customer feedback with military and security agencies—information that could refine China’s domestic monitoring systems and aid foreign intelligence gathering.
These privacy allegations have sparked unease among DeepSeek’s tens of millions of global users. Tech partners such as Amazon Web Services, Microsoft Azure, and Google Cloud have begun reviewing contractual clauses that might expose them to secondary sanctions if DeepSeek is officially deemed a national security risk. A memo circulated within one major cloud provider flagged potential liability under the U.S. Foreign Intelligence Surveillance Act (FISA) if American user data is transmitted to a foreign military without proper oversight.
The DeepSeek controversy underscores the growing intersection between AI development and geopolitical competition. As Washington tightens export controls on chips, it simultaneously faces pressure to maintain America’s edge in AI research—an area where private startups and academic institutions rely on the very hardware now restricted. Critics argue that overly broad exports bans could stifle innovation domestically, while supporters contend they are essential to prevent adversaries from fielding AI-enabled weapon systems.
Analysts note that China is not standing still. Beyond shell-company workarounds, Beijing has accelerated its own chip-fabrication efforts, funneling billions into domestic foundries capable of producing semiconductors that could one day rival Nvidia’s offerings. Meanwhile, research institutions linked to DeepSeek are said to be exploring bespoke chip designs optimized for AI workloads, raising alarms in Washington about potential leaps in China’s sovereign semiconductor capabilities.
Despite the gravity of the U.S. allegations, no new sanctions or export bans were announced at the State Department press briefing. Officials indicated that the evidence is still under review by multiple agencies, including the Departments of Commerce, Treasury, and Defense. Should DeepSeek be added to the Entity List maintained by the Bureau of Industry and Security, the firm would face an immediate cutoff from U.S.-origin technology—potentially crippling its training pipelines and data-processing operations.
In parallel, legal experts suggest DeepSeek could face civil penalties if found to have breached U.S. export law, with fines that scale according to the value of illegally procured equipment. At the same time, Congressional leaders are drafting amendments to the next National Defense Authorization Act, which could expand the definition of “national security technology” to include a broader array of AI tools and software—effectively closing existing loopholes exploited by DeepSeek and similar firms.
For now, DeepSeek continues to market its AI models globally, citing performance benchmarks that rival those of OpenAI and Meta. Yet with mounting scrutiny from Washington and potential investigations by several Southeast Asian governments into the shell-company networks, DeepSeek’s future as an international AI powerhouse appears increasingly uncertain.
(Source:www.reuters.com)
DeepSeek, headquartered in Hangzhou, captured global attention earlier this year when it unveiled AI reasoning models—DeepSeek-V3 and DeepSeek-R1—that it claimed matched or outperformed top U.S. industry offerings at a fraction of the cost. While those announcements fueled excitement across the tech world, U.S. authorities have since grown skeptical, concluding that the startup’s rapid ascent may owe more to covert support from the Chinese government and U.S.-origin technology than to breakthrough innovations.
Allegations of Military Collaboration
A senior State Department official, speaking on condition of anonymity, told reporters that DeepSeek “has willingly provided and will likely continue to provide support to China’s military and intelligence operations.” U.S. investigators point to internal procurement records in which DeepSeek appears more than 150 times in contracts linked to the People’s Liberation Army and affiliated defense research institutions. Those records, combined with statements from military procurement officers in Beijing, suggest the firm has supplied bespoke AI services to enhance surveillance, logistics planning, and other defense-related applications.
U.S. lawmakers reinforced these concerns when the House Select Committee on the Chinese Communist Party opened a parallel inquiry into DeepSeek’s corporate structure and technology transfers. A committee report detailed how founder Liang Wenfeng exercises de facto control over the company through a web of holding entities, raising questions about state influence behind DeepSeek’s ostensibly private-sector guise. That report also urged tighter export controls, warning of a “strategic surprise” if Beijing gains unfettered access to cutting-edge AI capabilities.
Circumventing Export Restrictions
Since 2022, U.S. policy has banned the sale of Nvidia’s H100 GPUs to mainland China, citing fears these chips could dramatically boost Beijing’s military-grade AI development. Despite those curbs, U.S. officials allege DeepSeek obtained “large volumes” of Nvidia’s high-performance chips by routing orders through a series of shell companies based in Malaysia, Singapore, and Vietnam. In one documented case, DeepSeek reportedly contracted with a small Singapore-based data center to host American-made GPUs, then accessed them remotely—an approach officials say exploits a loophole that still allows non-Chinese firms to rent U.S. chips in third-country facilities.
While DeepSeek has not been formally blacklisted by the U.S. government, State Department sources indicate that Commerce Department investigators are probing whether the firm violated the International Traffic in Arms Regulations (ITAR) or other export statutes. DeepSeek’s leadership denies any wrongdoing, insisting that all chips were acquired legally and that its models run on a mix of H800 series products—export-approved alternatives to the H100 line. Nvidia has likewise maintained that it adheres strictly to American export laws, asserting that any chips found in Southeast Asia were sold through non-China subsidiaries and without knowledge of their end-use.
Data Privacy and Global Scrutiny
In addition to export control concerns, Washington accuses DeepSeek of funneling user data to Beijing’s vast surveillance network. Under China’s 2017 National Intelligence Law, all domestic tech companies must comply with government requests for information. However, U.S. officials claim DeepSeek voluntarily shares detailed usage statistics, model-training logs, and even customer feedback with military and security agencies—information that could refine China’s domestic monitoring systems and aid foreign intelligence gathering.
These privacy allegations have sparked unease among DeepSeek’s tens of millions of global users. Tech partners such as Amazon Web Services, Microsoft Azure, and Google Cloud have begun reviewing contractual clauses that might expose them to secondary sanctions if DeepSeek is officially deemed a national security risk. A memo circulated within one major cloud provider flagged potential liability under the U.S. Foreign Intelligence Surveillance Act (FISA) if American user data is transmitted to a foreign military without proper oversight.
The DeepSeek controversy underscores the growing intersection between AI development and geopolitical competition. As Washington tightens export controls on chips, it simultaneously faces pressure to maintain America’s edge in AI research—an area where private startups and academic institutions rely on the very hardware now restricted. Critics argue that overly broad exports bans could stifle innovation domestically, while supporters contend they are essential to prevent adversaries from fielding AI-enabled weapon systems.
Analysts note that China is not standing still. Beyond shell-company workarounds, Beijing has accelerated its own chip-fabrication efforts, funneling billions into domestic foundries capable of producing semiconductors that could one day rival Nvidia’s offerings. Meanwhile, research institutions linked to DeepSeek are said to be exploring bespoke chip designs optimized for AI workloads, raising alarms in Washington about potential leaps in China’s sovereign semiconductor capabilities.
Despite the gravity of the U.S. allegations, no new sanctions or export bans were announced at the State Department press briefing. Officials indicated that the evidence is still under review by multiple agencies, including the Departments of Commerce, Treasury, and Defense. Should DeepSeek be added to the Entity List maintained by the Bureau of Industry and Security, the firm would face an immediate cutoff from U.S.-origin technology—potentially crippling its training pipelines and data-processing operations.
In parallel, legal experts suggest DeepSeek could face civil penalties if found to have breached U.S. export law, with fines that scale according to the value of illegally procured equipment. At the same time, Congressional leaders are drafting amendments to the next National Defense Authorization Act, which could expand the definition of “national security technology” to include a broader array of AI tools and software—effectively closing existing loopholes exploited by DeepSeek and similar firms.
For now, DeepSeek continues to market its AI models globally, citing performance benchmarks that rival those of OpenAI and Meta. Yet with mounting scrutiny from Washington and potential investigations by several Southeast Asian governments into the shell-company networks, DeepSeek’s future as an international AI powerhouse appears increasingly uncertain.
(Source:www.reuters.com)
Homepage
Send to a friend
Printable version
Share
