On April 27, 2026, Chinese AI company DeepSeek’s release of the V4 new model has sparked great enthusiasm among mainland Chinese media. At the same time, the White House and the State Department of the United States have intensively acted, issuing memorandums and global diplomatic cables, pointing directly at Chinese companies represented by DeepSeek for allegedly using “distillation” methods to systematically steal US core AI technology. Experts believe that this controversy has far exceeded the scope of commercial competition and is evolving into a deep-seated game concerning national security and global AI governance.
DeepSeek unveiled the new generation V4 model on April 24, drawing high attention in China. Many voices claim it has “surpassed the United States.” However, several media outlets and experts believe that such statements are overly optimistic.
DeepSeek claims that its V4 series has made progress in encoding, reasoning, and intelligent body tasks, emphasizing cost advantages. At the same time, they admit that they are still behind the most advanced models in the US by about 3 to 6 months on some key indicators.
However, Lin Junyang, the former technical lead of the chatbot “Qwen” developed by Alibaba’s Alibaba Cloud, who left Alibaba in early March, stated at an AI forum at Beijing Tsinghua University in January this year that the likelihood of any Chinese company surpassing US AI giants like Google DeepMind and OpenAI within the next 3 to 5 years is less than 20%. “This is a very optimistic estimate because the computing resources in the US are 10 to 100 times greater than in China.”
Tang Jie, co-founder and Chief AI Scientist of Z.ai, also mentioned at the forum, “The gap between China and the US may actually be widening because the US has many AI models that have not been publicly released.”
Professor Sun Guoxiang from the Department of International Affairs and Business at Nanhua University in Taiwan, in an interview with Dajiyuan, pointed out that from the perspective of public opinion, DeepSeek does have a symbolic significance of “breaking through the blockade,” but in terms of real AI capabilities, “it is still premature to say it has completely surpassed the United States.”
“China’s strengths lie in low-cost diffusion, fast engineering, localization, and domestic alternatives, but the strengths of the United States still lie in cutting-edge models, computing power, ecosystems, basic research, chips, and cloud platforms,” he said.
Su Ziyun, Director of the Strategy and Resources Department of the Taiwan Institute for National Defense Security, told Dajiyuan, “The CCP is now using methods similar to the Great Leap Forward movement in steel production to develop artificial intelligence and chips, and they will soon hit a ceiling.”
First, Chinese AI companies are currently not profitable, whether it’s DeepSeek, Tencent, or Baidu’s AI, they are all in a state of loss mainly due to a lack of a comprehensive charging mechanism. Many usage scenarios consume a large amount of computing power and electricity but do not bring in revenue. AI generated content like the previously promoted “Snow Mountain Sauce Plate Duck Saves Fox” may spark discussions but does not generate actual commercial value.
Secondly, there are limitations to hardware. The current method involves compensating for computing power with electricity using inefficient chips, resulting in huge power consumption.
Thirdly, Chinese AI companies currently use so-called “advanced technology,” which is actually outcomes stolen from algorithms of US AI companies. Su Ziyun pointed out, “This is like peeking at simulated exam questions, which may raise security concerns.”
Following the release of DeepSeek V4, the US government has escalated its accusations and warnings against Chinese AI companies. On April 23, the White House issued a memorandum accusing China of “industrial-scale” theft of American AI technology.
As early as January 2025, DeepSeek released the R1 and V3 versions, with their low-cost gimmick causing shock in the US tech industry. In February of the same year, OpenAI submitted a memorandum to the US Congress, stating that actors from China and Russia were continuing complex distillation activities.
In February of this year, US AI company Anthropic accused three leading Chinese AI companies – DeepSeek, Moonshot, and MiniMax – of launching distillation attacks on their models.
Distillation refers to using outputs of larger, more expensive AI models to train smaller AI models, aiming to reduce the cost of training powerful new AI tools.
Recently, Reuters cited a diplomatic cable showing that the US State Department has ordered warnings to be issued globally to provoke broad attention to Chinese companies like DeepSeek stealing US AI lab intellectual property through “distillation” techniques.
Sun Guoxiang analyzed, the key to this controversy is that the US has elevated it from commercial infringement to a national security issue. “US policies have had various impacts on the CCP, including technological pressure, economic pressure, and international pressure.”
“In the future, the global AI market may split into two trust systems: one led by the US, emphasizing security reviews and intellectual property protection, and the other promoted by the CCP, focusing on low-cost open source, localization, and rapid deployment,” he said.
On April 25, US President Trump stated that the US aims to defeat China (CCP) in various industry sectors. Currently, the US is leading in cryptocurrency and artificial intelligence (AI) fields.
He reiterated the White House cryptocurrency advisors’ support for the Digital Asset Market Clarity Act and criticized banking lobbyists obstructing the progress of the bill. The legislation is considered a primary policy goal for the cryptocurrency industry.
Sun Guoxiang pointed out that this statement not only serves as a political declaration but also embodies triple strategic implications: AI represents productivity and military power, cryptocurrency involves the dollar system, and the overall competition is a battle of institutional models.
Su Ziyun stated, “Even the US Indo-Pacific Command openly stated that using blockchain methods for cryptocurrencies will enhance the overall communication security for the US military in the event of a war with China (CCP). In terms of national security, including data security in military operations, artificial intelligence, and advanced digital chips are very crucial.”