Politics

US warns diplomats over China’s alleged AI theft and new DeepSeek model adaptation

Monday, 27 April 2026 6:01AM UTC

The US State Department has instructed its embassies worldwide to raise concerns over Chinese efforts to extract and replicate American AI technology, signalling a widening campaign amid escalating tensions with China, with new developments from Chinese firm DeepSeek heightening the rivalry.

The US State Department has told diplomats to raise concerns around the world over what it says are Chinese efforts to extract know-how from American artificial intelligence systems, a sign that Washington is widening its campaign against alleged AI theft as tensions with Beijing deepen. According to a diplomatic cable seen by Reuters, the instruction was sent on Friday to embassies and consulates and specifically referred to concerns about the “extraction and distillation” of US AI models. The cable also said a separate message had been sent to Beijing.

The warning comes as the White House has begun making similar allegations more openly. A memo from Michael Kratsios, the director of the White House Office of Science and Technology Policy, accused China-backed actors of carrying out “industrial-scale” efforts to steal and replicate frontier AI systems, using proxy accounts and jailbreak techniques to evade detection. Axios reported that the administration plans to share intelligence with US AI companies so they can strengthen defences against such tactics.

DeepSeek has become the clearest symbol of that rivalry. The Hangzhou-based startup, which jolted global markets last year with a low-cost model that appeared to narrow the gap with leading US systems, has now released a preview of its V4 model, according to Reuters and AP. The new version is adapted for Huawei chips, underscoring China’s push to reduce dependence on American hardware. AP said the company is offering open-source variants and claims the system improves reasoning, knowledge and autonomous workflow capabilities.

The US allegations are not limited to DeepSeek. Reuters said the State Department cable also names Moonshot AI and MiniMax, while OpenAI has separately warned lawmakers that DeepSeek was trying to imitate its ChatGPT models for training purposes. The cable argued that models developed through unauthorised distillation can look competitive on selected benchmarks while failing to reproduce the broader performance of the original systems, and it said such methods can also strip away safety controls.

Beijing has rejected the charges. The Chinese embassy in Washington told Reuters that the allegations were “groundless” and amounted to deliberate attacks on China’s AI progress. DeepSeek has also said in the past that its earlier models were trained on data gathered through web crawling rather than on synthetic outputs from OpenAI. The dispute lands just weeks before a planned visit by Donald Trump to Beijing, raising the risk that the AI fight becomes another flashpoint in a broader tech confrontation between the world’s two largest economies.

Source Reference Map

Inspired by headline at: ^[1]

Sources by paragraph:

Paragraph 1: ^[2], ^[1]
Paragraph 2: ^[3], ^[4]
Paragraph 3: ^[5], ^[6], ^[1]
Paragraph 4: ^[2], ^[4], ^[1]
Paragraph 5: ^[1], ^[2], ^[3]

Source: Noah Wire Services

More on this

https://www.rappler.com/technology/us-state-department-orders-global-warning-ai-theft-deepseek-chinese-firms/ - Please view link - unable to able to access data
https://www.axios.com/2026/04/23/us-china-ai-theft-distillation - The Trump administration has accused China-backed operatives of conducting 'industrial-scale' operations to steal and replicate cutting-edge American artificial intelligence (AI) models. According to a memo from Michael Kratsios, Director of the White House Office of Science and Technology Policy, the actors—primarily based in China—used proxy accounts to bypass detection and manipulate AI systems in order to extract proprietary information. This development intensifies the ongoing U.S.-China AI rivalry and adds tension ahead of President Trump's upcoming visit to Beijing, where critical discussions on economic relations are expected. The U.S. has previously accused China of intellectual property theft, particularly through cyber-espionage tactics. While acknowledging that AI capabilities derived from stolen models may not be durable, Kratsios emphasized the administration’s intent to share intelligence with U.S. AI companies and support the development of stronger defenses. ([axios.com](https://www.axios.com/2026/04/23/us-china-ai-theft-distillation?utm_source=openai))
https://www.techradar.com/pro/security/we-will-be-taking-action-to-protect-american-innovation-white-house-accuses-china-of-systematically-extracting-american-ai-technologies - The White House has accused China of orchestrating a large-scale, coordinated effort to extract sensitive American artificial intelligence (AI) technologies. According to a memo from Michael Kratsios, Director of the White House Office of Science and Technology Policy, foreign entities, primarily based in China, are using tens of thousands of proxy accounts and jailbreak techniques to bypass defenses and access proprietary American AI systems. These efforts are described as 'industrial-scale' and are characterized as a direct threat to U.S. innovation. The U.S. government has pledged to take retaliatory actions to protect its AI sector, framing this behavior as malicious exploitation rather than innovation. The memo comes ahead of a high-stakes meeting between U.S. President Donald Trump and Chinese President Xi Jinping in Beijing, and just as Congress is preparing expansive tech export control legislation to curtail China’s access to advanced American technologies like high-performance chips. China has rejected the allegations, emphasizing its commitment to intellectual property rights. Despite restrictions, China has made significant progress in AI development, surprising U.S. developers with advanced models like Deepseek and OpenClaw, narrowing the expected technological gap between the two nations. ([techradar.com](https://www.techradar.com/pro/security/we-will-be-taking-action-to-protect-american-innovation-white-house-accuses-china-of-systematically-extracting-american-ai-technologies?utm_source=openai))
https://www.pcgamer.com/software/ai/white-houses-claim-that-china-is-engaged-in-deliberate-industrial-scale-campaigns-to-distil-us-frontier-ai-systems-called-pure-slander-by-chinese-embassy/ - A recent White House memo, revealed by the Financial Times, alleges that Chinese entities are systematically engaging in 'industrial-scale campaigns' to distill cutting-edge U.S. AI systems. AI distillation refers to creating a new model by training it on the outputs of a more advanced one, potentially replicating proprietary technology. The White House reportedly plans to share intelligence with American AI firms to help them defend against such activities. Companies like Anthropic have previously accused Chinese firms DeepSeek, Moonshot, and MiniMax of such actions. Michael Kratsios, a senior U.S. science and technology adviser, also claimed that Chinese operations are using numerous proxy accounts and jailbreaking techniques to bypass detection and extract proprietary information. The U.S. government may take steps to hold these foreign actors accountable. In response, the Chinese embassy in Washington denied the allegations, calling them 'pure slander' and emphasizing China's commitment to intellectual property rights. The issue gains urgency as Chinese AI companies, such as DeepSeek with its newly launched DeepSeek-V4 model, continue advancing rapidly, increasing U.S. concerns over IP theft and competitive disadvantage in the AI landscape. ([pcgamer.com](https://www.pcgamer.com/software/ai/white-houses-claim-that-china-is-engaged-in-deliberate-industrial-scale-campaigns-to-distil-us-frontier-ai-systems-called-pure-slander-by-chinese-embassy/?utm_source=openai))
https://apnews.com/article/d2ed33f2521917193616e061674d5f92 - Chinese AI startup DeepSeek has unveiled its highly anticipated V4 AI model, intensifying the growing technological rivalry between China and the United States. The V4 rollout includes two open-source versions—'Pro' and 'Flash'—that significantly improve knowledge, reasoning, and agentic (autonomous workflow) capabilities. This update is notable for its use of Huawei's chips, reducing dependence on U.S. hardware. Compared to U.S. counterparts, DeepSeek claims V4 Pro Max competes strongly with OpenAI's GPT-5.2 and Google's Gemini 3.0-Pro, though it trails slightly behind GPT-5.4 and Gemini 3.1-Pro. The models feature a 1 million token context window, a substantial advancement from V3’s 128,000 tokens. DeepSeek gained global attention in early 2025 with its R1 model, known for its cost-effective reasoning capabilities, signaling China's AI strides. While the company promotes openness in allowing developers access to its architecture, it faces serious allegations from OpenAI and Anthropic for purportedly distilling U.S. models to build their own. U.S. officials, including Michael Kratsios, have echoed concerns about Chinese firms exploiting American AI. Nonetheless, analysts see V4 as a pivotal move for China's AI ambitions, though independent benchmarks are needed to verify its performance. ([apnews.com](https://apnews.com/article/d2ed33f2521917193616e061674d5f92?utm_source=openai))
https://www.tomshardware.com/tech-industry/artificial-intelligence/deepseek-launches-1-6-trillion-parameter-v4-on-huawei-chips-as-u-s-escalates-ai-theft-accusations - DeepSeek on Friday released a preview of its V4 large language model, the Hangzhou-based startup's most powerful to date, with 1.6 trillion parameters and a 1 million token context window. The model is the first major frontier release optimized for Huawei's Ascend AI processors rather than Nvidia hardware, and it arrived on the same day Reuters reported that the U.S. State Department had sent a diplomatic cable to embassies worldwide instructing staff to warn foreign governments about alleged

Noah Fact Check Pro

The draft above was created using the information available at the time the story first emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed below. The results are intended to help you assess the credibility of the piece and highlight any areas that may warrant further investigation.

Freshness check

Score: 8

Notes: The article reports on a diplomatic cable dated April 24, 2026, instructing U.S. diplomatic and consular posts to raise concerns about Chinese companies' alleged extraction and distillation of U.S. AI models. ([gmanetwork.com](https://www.gmanetwork.com/news/scitech/technology/985270/us-state-dept-orders-global-warning-about-alleged-china-ai-thefts-by-deepseek-others/story/?utm_source=openai)) This information is corroborated by multiple reputable sources, including Reuters and The Straits Times, published on April 25, 2026. ([straitstimes.com](https://www.straitstimes.com/world/united-states/us-state-dept-orders-global-warning-about-alleged-china-ai-thefts-by-deepseek-others?utm_source=openai)) The content appears original and timely, with no evidence of prior publication or significant overlap with other sources.

Quotes check

Score: 7

Notes: The article includes direct quotes from the diplomatic cable, such as "concerns over adversaries' extraction and distillation of U.S. A.I. models." ([gmanetwork.com](https://www.gmanetwork.com/news/scitech/technology/985270/us-state-dept-orders-global-warning-about-alleged-china-ai-thefts-by-deepseek-others/story/?utm_source=openai)) These quotes are consistent with those found in other reputable sources, indicating they are not fabricated. However, the exact wording of the cable is not independently verifiable, as the full text has not been publicly released. This limits the ability to fully authenticate the quotes.

Source reliability

Score: 9

Notes: The article is based on a Reuters report, a major news organisation known for its credibility and independence. ([gmanetwork.com](https://www.gmanetwork.com/news/scitech/technology/985270/us-state-dept-orders-global-warning-about-alleged-china-ai-thefts-by-deepseek-others/story/?utm_source=openai)) The information is corroborated by other reputable sources, including The Straits Times and The Economic Times, published on April 25, 2026. ([straitstimes.com](https://www.straitstimes.com/world/united-states/us-state-dept-orders-global-warning-about-alleged-china-ai-thefts-by-deepseek-others?utm_source=openai)) The sources are independent and not affiliated with any parties mentioned in the article, enhancing the reliability of the information.

Plausibility check

Score: 8

Notes: The claims of the U.S. State Department warning about alleged AI theft by Chinese firms align with recent reports from other reputable sources, including The Straits Times and The Economic Times, published on April 25, 2026. ([straitstimes.com](https://www.straitstimes.com/world/united-states/us-state-dept-orders-global-warning-about-alleged-china-ai-thefts-by-deepseek-others?utm_source=openai)) The article provides specific details, such as the mention of Chinese AI firms DeepSeek, Moonshot AI, and MiniMax, which are consistent with other reports. The language and tone are appropriate for the topic and region, and there are no signs of excessive or off-topic detail. The report appears plausible and consistent with other reputable sources.

Overall assessment

Verdict (FAIL, OPEN, PASS): PASS

Confidence (LOW, MEDIUM, HIGH): HIGH

Summary: The article presents timely and original information about the U.S. State Department's warning regarding alleged AI theft by Chinese firms. The claims are corroborated by multiple reputable sources, and the content is consistent with standard journalistic practices. While the full text of the diplomatic cable has not been publicly released, the information provided is plausible and supported by independent verification. Therefore, the article passes the fact-check with high confidence.

AI
US-China relations
intellectual property theft