WASHINGTON (Reuters) – The U.S. State Department has issued a global alert for widespread efforts by Chinese companies, including AI startup DeepSeek, to steal intellectual property from U.S. artificial intelligence labs, according to a diplomatic cable seen by Reuters.
The cable, dated Friday and addressed to diplomats and consulates around the world, instructs diplomatic staff to speak with their foreign counterparts about “concerns about the extraction and distillation of the USAI model by adversaries.”
“Another cancellation request and message was sent to the Chinese government seeking negotiations with China,” the document said.
Distillation is the process of using the output from larger, more expensive AI models to train smaller AI models as part of an effort to reduce the cost of training powerful new AI tools.
The White House made similar accusations this week, but the cables were not previously reported. The State Department did not immediately respond to a request for comment.
OpenAI warned U.S. lawmakers that DeepSeek was targeting the ChatGPT maker and the nation’s leading AI companies to clone its models and use them for its own training, Reuters reported in February.
China rejects accusations
The Chinese embassy in Washington reiterated its position on Friday that the accusations are baseless.
“Claims that Chinese companies are stealing U.S. AI intellectual property are baseless and a deliberate attack on China’s development and the progress of the AI industry,” it said in a statement to Reuters.
DeepSeek, which surprised the world last year with its low-cost AI model, on Friday announced a preview of its long-awaited new model called V4, which is adapted to Huawei’s chip technology, underscoring China’s growing autonomy in this field.
DeepSeek also did not immediately respond to a request for comment. The company has previously said its V3 model uses data that occurs naturally and is collected through web crawling, and does not intentionally use synthetic data generated by OpenAI.
Many Western countries and some Asian governments have banned the use of DeepSeek by their agencies and officials, citing data privacy concerns. Nevertheless, DeepSeek’s models have always been most popular on international platforms hosting open source models.
The State Department Cable said its purpose was to “warn the risks of the use of AI models derived from U.S. proprietary AI models and lay the groundwork for potential follow-up and assistance by the U.S. government.”
It also mentioned Chinese AI companies Moonshot AI and MiniMax (0100.HK), but neither responded to requests for comment.
“AI models developed from covert and unauthorized distillation campaigns allow foreign attackers to release products that appear to perform as well on some benchmarks at a fraction of the cost, but cannot reproduce the full performance of the original system,” the Telegram said.
It added that the campaign “intentionally removes security protocols from the resulting models and reinstates mechanisms that ensure that AI models are ideologically neutral and truth-seeking.”
The White House accusation and cable were released just weeks before US President Donald Trump is scheduled to visit Chinese President Xi Jinping in Beijing. Although the détente brokered last October had eased tensions, tensions in the long-running technology war between the rival superpowers are likely to escalate.
