US State Dept orders global push to warn of DeepSeek AI ‘distillation’ thefts
The US State Department has reportedly mandated a global expansion of an extensive initiative in response to moves by Chinese companies, including AI startup DeepSeek.
The primary motive behind this revelation is concerns that these entities aim to capture intellectual property from US artificial intelligence labs.
A cable sent to diplomatic and consular posts worldwide calls on officials to embark on missions to speak with overseas peers regarding “concerns over adversaries” extraction and distillation of US AI models.”
The document reads: “A separate demarche request and message has been sent to Beijing to be raised with China.”
For those unversed, distillation is the process of training smaller AI models using output from larger, more expensive models in an effort to mitigate the costs of training a powerful new AI tool.
The White House has reportedly made similar accusations, but the specific cable has not been previously reported. In this regard, the Chinese Embassy in Washington expressed its stance that all these accusations are baseless.
According to Reuters, it said in a statement: “The allegations that Chinese entities are stealing American AI intellectual property are groundless and are deliberate attacks on China’s development and progress in the AI industry.”
The Chinese AI startup, DeepSeek, which surprised the world with a low-cost model last year, has introduced a preview of its highly anticipatedV4 series, released this Friday.
Previously, DeepSeek maintained that its V3 model used data collected through web scraping and denied intentionally using simulated data generated by OpenAI.
Given the gravity of the situation, many Western and some Asian governments have banned their institutions and officials from using DeepSeek, for the sake of personal data security. It is pertinent to note that DeepSeek’s models are considered among the most widely used open-source models on international platforms.
The V4 Pro version surpasses other open-source models in world-knowledge, trailing only Google’s Gemini 3.1 Pro.
The company also launched V4 Flash, a more affordable version designed to incorporate real-world user feedback. DeepSeek has faced repeated accusations from the US and its allies regarding improper and illegal conduct.
The State Department cable further proposed the idea behind the current revelation is to “warn of the risks of utilizing AI models distilled from US proprietary AI models and lay the groundwork for potential follow-up and outreach by the US government.”
Imitation models are built using unauthorized data to mimic original systems; they appear to perform well on specific tests at a much lower cost but lack the overall quality of the original.
These campaigns intentionally remove built-in security protocols, making the resulting models less safe. The process undoes mechanisms designed to keep the AI ideologically neutral and truth-seeking, potentially allowing the models to produce biased or misinformation-heavy content.
Additionally, these “surreptitious” methods allow foreign entities to bypass the massive research and development costs required to build high-level AI from scratch.
Nonetheless, the White House accusations come just weeks before US President Trump is scheduled to visit Chinese President Xi Jinping in Beijing.
