US State Department orders global warning about alleged AI theft by DeepSeek, other Chinese firms

In a diplomatic cable seen by Reuters, the US State Department has called for a worldwide campaign to draw international attention to alleged attempts by China-based companies, including DeepSeek, which develops artificial intelligence technology, to steal intellectual property from US AI laboratories.

The cable, released on Friday and sent to diplomatic missions around the world, asks diplomats to engage with their host countries and discuss “concerns over adversaries’ extraction and distillation of US AI models.”

“A separate demarche request and message have been sent to Beijing to raise these concerns with China,” the document added.

Distillation refers to the process by which smaller AI models are trained using the output generated by larger models. This allows companies to reduce the costs associated with developing such systems.

The White House has previously made similar allegations against China.

In a letter to Congress last month, OpenAI stated that DeepSeek was attempting to extract information from the company and other leading US AI firms to develop its own models.

The Chinese Embassy in Washington on Friday reaffirmed its position that the claims are unfounded.

In a statement to Reuters, it said accusations that Chinese entities are stealing US AI intellectual property are baseless and amount to deliberate attempts to undermine China’s progress in the artificial intelligence sector.

DeepSeek, whose low-cost AI model drew global attention last year, introduced a preview of its much-anticipated new model, V4, on Friday. The model is designed to work with Huawei chip technology, highlighting China’s increasing self-reliance in the field.

The company did not immediately respond to requests for comment.

Previously, DeepSeek stated that its V3 model was trained on naturally sourced data gathered through web crawling and did not intentionally rely on synthetic data produced by OpenAI.