U.S. Navy Bans DeepSeek AI Over Security Risks

The U.S. Navy has banned DeepSeek AI over security concerns, citing risks tied to Chinese data policies and potential espionage links as the AI model has surged in global adoption.

The U.S. Navy has issued an internal directive prohibiting its personnel from using DeepSeek AI, a China-based artificial intelligence platform, citing concerns over security risks, potential data privacy violations, and ethical issues tied to its origins.

The ban applies to both work-related and personal use, signaling growing U.S. government scrutiny over AI systems developed in China.

The directive, which was sent on Friday via the Operational Navy (OpNav) network, explicitly instructs members to “refrain from downloading, installing, or using the DeepSeek model in any capacity.”

A U.S. Navy spokesperson confirmed the authenticity of the email to CNBC, stating that the decision aligns with the Department of the Navy’s Chief Information Officer’s generative AI policy.

The warning was based on an advisory from the Naval Air Warfare Center Division Cyber Workforce Manager, underscoring the security risks that DeepSeek AI may pose.

Related: DeepSeek Drops Another OpenAI-Buster With Janus Multimodal Models, Outpacing DALL-E 3

The move mirrors previous U.S. government actions against Chinese-owned digital platforms, such as the restrictions placed on TikTok for government employees over concerns that data collected by the app could be accessed by Chinese authorities. However, the Navy’s decision comes at a time when DeepSeek AI’s influence is rapidly growing.

The platform recently surpassed OpenAI’s ChatGPT in Apple’s App Store rankings, raising alarm among policymakers about how much user data is being transmitted to China and whether the AI model could be leveraged for espionage.

DeepSeek’s Unprecedented AI Growth Raises Red Flags

DeepSeek AI’s rise to prominence has been unusually swift. The company’s latest large-scale language model, DeepSeek R1, was introduced as a free, open-source AI system, claiming to match or outperform OpenAI’s o1 model in reasoning tasks while being significantly cheaper to develop.

DeepSeek executives have stated that R1 was trained on only 2,048 Nvidia H800 GPUs, a China-approved version of the H100 chips that the U.S. has placed under export controls.

The company also reported that the entire model was built with a budget of just $6 million, an astonishingly low figure compared to the hundreds of millions of dollars typically required to train models of similar sophistication.

Related: Alibaba’s New Qwen 2.5-Max Model Takes on DeepSeek in AI Benchmarks

However, AI industry leaders and cybersecurity experts have raised doubts about these claims. Alexandr Wang, CEO of Scale AI, said in a CNBC interview that DeepSeek’s hardware access is likely far greater than publicly disclosed.

“DeepSeek has about 50,000 Nvidia H100 GPUs. They can’t talk about it because it violates U.S. export controls,” Wang stated, suggesting that DeepSeek may have acquired restricted U.S. hardware through unofficial supply chains or smuggling operations.

If true, this would indicate that DeepSeek has been stockpiling high-performance GPUs despite U.S. restrictions, allowing it to train its models at a scale comparable to OpenAI, Google, and Anthropic—companies that collectively spend billions of dollars on AI model training.

Espionage Allegations and Unauthorized Use of OpenAI Data

Beyond hardware concerns, DeepSeek is also under investigation for potentially using stolen AI training data from OpenAI.

Both Microsoft and OpenAI have launched internal probes to determine whether DeepSeek used data scraping techniques to extract model outputs and replicate OpenAI’s research.

A security review flagged unusual API activity from a China-linked research group in late 2024, raising questions about how DeepSeek could have fast-tracked the development of R1 with minimal public training data.

The most alarming claim linking DeepSeek to unauthorized AI data acquisition is tied to the mysterious death of OpenAI researcher Suchir Balaji. Balaji, a 26-year-old AI specialist who worked on OpenAI’s training data pipelines, was found dead in his San Francisco apartment on November 26, 2024. Authorities ruled his death a suicide within 40 minutes of arrival, leaving little room for further investigation.

Investigative journalist George Webb, who has documented alleged AI-related espionage cases, suggested that Balaji may have been targeted for his knowledge of OpenAI’s data training methods.

“Balaji was found dead in his San Francisco apartment, and within 40 minutes, it was ruled a suicide. No real investigation, no effort to connect the dots,” Webb told CNBC.

Webb argues that DeepSeek’s AI model may incorporate proprietary OpenAI research, potentially obtained through illicit data scraping, insider leaks, or cyber intrusions. Microsoft is reportedly investigating whether OpenAI data was compromised, though the company has not publicly confirmed these allegations.

Growing Military and Government Response to AI Security Threats

The U.S. Navy’s ban on DeepSeek AI is part of a broader military effort to limit the use of foreign AI systems with unknown data governance practices. The military has long exercised caution with AI applications, previously restricting the use of OpenAI’s ChatGPT for security-sensitive tasks. The Navy’s move may signal upcoming federal actions that extend beyond military personnel.

President Donald Trump, who returned to office last Monday, weighed in on the DeepSeek controversy, stating “This should be a wake-up call for America’s tech companies.”

The Trump administration has already taken steps to bolster domestic AI development through a newly announced initiative called the Stargate Project, a joint venture between OpenAI, Oracle, and SoftBank aimed at building AI infrastructure in the U.S. to counter China’s advancements.

Meanwhile, David Sacks, Trump’s AI and crypto policy advisor, acknowledged the growing challenge posed by DeepSeek:

“The AI race will be very competitive, and we can’t be complacent,” he said, emphasizing the need for greater U.S. investment in AI security.

European Regulators Investigate DeepSeek for Data Privacy Violations

Concerns about DeepSeek’s operations are not confined to the U.S. Italy’s data protection authority, Garante, has launched a formal investigation into whether DeepSeek violated the European Union’s General Data Protection Regulation (GDPR) by transferring European user data to servers in China.

DeepSeek has until February 17 to respond to the inquiry, and failure to comply could lead to fines or restrictions on the company’s operations within the European Union.

AI Export Controls and the Future of AI Security Policy

The DeepSeek controversy is intensifying calls for stricter AI export controls in the U.S. and Europe. Lawmakers in Washington and Brussels are discussing new measures, including a mandatory chip registry that would require AI firms to report their use of high-performance GPUs to prevent large-scale acquisitions of restricted hardware.

Alexandr Wang emphasized the urgency of expanding U.S. AI infrastructure in response to China’s progress. “The U.S. needs a lot more computational capacity. We are talking about an infrastructure challenge at a national level. If we don’t act fast, China will not just be competing with us—they will be leading,” he said.

The Navy’s ban on DeepSeek AI is the latest in a series of measures aimed at controlling China’s influence in AI development, but the broader challenge remains.

DeepSeek’s sudden rise, questionable training methods, and opaque ties to the Chinese government raise serious questions about the future of AI security and AI safety — topics that will likely remain a major point of contention in global technology policy discussions.

Markus Kasanmascheff
Markus Kasanmascheff
Markus has been covering the tech industry for more than 15 years. He is holding a Master´s degree in International Economics and is the founder and managing editor of Winbuzzer.com.

Recent News

0 0 votes
Article Rating
Subscribe
Notify of
guest
0 Comments
Newest
Oldest Most Voted
Inline Feedbacks
View all comments
0
We would love to hear your opinion! Please comment below.x
()
x