As global tensions around artificial intelligence escalate, a new player has landed in the crosshairs of Washington’s export control watchdogs — and it’s not a household name like Huawei or Tencent.
It’s DeepSeek, an AI startup based in Hangzhou, China, that made headlines earlier this year after claiming its models could rival — or outperform — OpenAI’s GPT-4, at a fraction of the cost.
But according to a senior US State Department official speaking to Reuters, DeepSeek’s ascent may have come with high geopolitical costs. The official alleges that DeepSeek is providing support to China’s military and intelligence operations, while also attempting to evade US export controls on advanced semiconductor technology.
These claims, if substantiated, could mark a turning point in how Washington regulates AI innovation tied to China — and raise new questions about the role of startups in the evolving AI cold war.
From AI Darling to National Security Concern
DeepSeek stunned the tech world in January after boasting that its DeepSeek-V3 and DeepSeek-R1 models matched the performance of top US counterparts — all for a reported $5.58 million in training costs.
Silicon Valley engineers were impressed. But US officials are skeptical.
“This goes well beyond open-source access,” the senior official said, alleging that DeepSeek is actively supplying China’s People’s Liberation Army (PLA) with AI technologies and is cited in over 150 Chinese military procurement records.
What’s more, the company is said to be sharing user data and statistics with Beijing’s surveillance system — a violation that could raise alarm for DeepSeek’s tens of millions of global users, including those in the US.
The Nvidia Loophole: Export Controls Under Strain
At the heart of the controversy is Nvidia’s H100 — the gold-standard AI chip that’s currently restricted from sale to China due to concerns over its potential use in military-grade AI development.
DeepSeek, the official says, attempted to sidestep these restrictions by:
- Setting up shell companies in Southeast Asia
- Remotely accessing Nvidia-powered data centers located outside of China
- Allegedly acquiring H100 chips through third parties, despite the ban
Singaporean authorities have already charged three men in a fraud case reportedly tied to the shipment of advanced chips to DeepSeek, while Malaysia is investigating whether an unnamed Chinese company is training large language models on US chips without authorization.
Nvidia has denied any involvement, stating:
“We do not support parties that have violated US export controls… We are effectively out of the China data center market.”
The company added that their internal review suggests DeepSeek used H800 chips, a less powerful variant still permitted under certain restrictions — not the banned H100s.
What This Means for Global AI Regulation
The DeepSeek allegations come as the US steps up enforcement of its AI and semiconductor export rules. Washington has made clear that it intends to choke off access to high-end chips that could help Beijing accelerate military AI development.
But the DeepSeek case shows how digital workarounds and jurisdictional loopholes may be undermining those efforts. As long as Chinese firms can access restricted chips through foreign data centers or third-party proxies, the effectiveness of US sanctions will be limited.
So far, DeepSeek has not been placed on a US entity list, and no formal sanctions have been announced. But pressure is mounting — both from national security hawks and from industry insiders who fear a shadow AI supply chain is already operating at scale.
Big Tech, Small Company — Big Implications
DeepSeek’s meteoric rise — from obscurity to global competitor — highlights a broader challenge for US policymakers. In today’s AI landscape, it’s not just state giants like Baidu or Alibaba that pose strategic risks.
Startups, fueled by opaque capital flows and cross-border cloud access, can achieve outsized influence in months. And if even one of them is helping a military adversary train smarter, faster systems, the consequences could ripple far beyond the tech sector.
Final Words
The story of DeepSeek isn’t just about one company. It’s about the collision between innovation and geopolitics — and the uncomfortable truth that the AI race may not be won by the most open or most ethical players.
As the US weighs next steps, the tech world should pay attention. Because in this new era of AI power projection, data, chips, and code are the new strategic weapons — and startups like DeepSeek are already in the arena.
