The US State Department has ordered a global push to bring attentionto what it says are widespread efforts by Chinese companies,including AI startup DeepSeek, to steal intellectual property fromU.S. artificial intelligence labs, according to a diplomatic cableseen by Reuters.
The cable said itspurpose was to “warn of the risks of utilizing AI modelsdistilled from US proprietary AI models, and lay the groundwork forpotential follow-up and outreach by the US government.”
Distillation is theprocess of training smaller AI models using output from larger,more expensive ones as part of an effort to lower the costs oftraining a powerful new AI tool.
DeepSeek, theChinese startup whose low-cost AI model stunned the world last year,on Friday launched a preview of a highly anticipated new modeladapted for Huawei chip technology, underlining China’s growingautonomy in the sector.
The StateDepartment, DeepSeek and the Chinese Embassy in Washington did notimmediately respond to a request for comment.
The cable alsomentioned Chinese AI firms Moonshot AI and MiniMax. Neither companyimmediately responded to a request for comment.
This week, the WhiteHouse made similar accusations, which the Chinese EmbassyinWashington called “baseless allegations,” adding thatBeijing “attaches great importance to the protection ofintellectual property rights.”
The cable, datedFriday and sent to diplomatic and consular posts around the world,instructs diplomatic staff to speak to their foreign counterpartsabout “concerns over adversaries’ extraction and distillation ofUS A.I. models.”
“A separatedemarche request and message has been sent to Beijing for raisingwith China,” the document states.
The cable, which hasnot been previously reported, signals the Trump administration istaking seriously the growing concerns about Chinese distillation ofUS AI models.
“AI modelsdeveloped from surreptitious, unauthorized distillation campaignsenable foreign actors to release products that appear to performcomparably on select benchmarks at a fraction of the cost but do notreplicate the full performance of the original system,” thecable said, adding that the campaigns also “deliberately stripsecurity protocols from the resulting models and undo mechanisms thatensure those AI models are ideologically neutral and truthseeking.”
OpenAI warned U.S.lawmakers that Chinese artificial intelligence startup DeepSeek wastargeting the ChatGPT maker and the nation’s leading AI companies toreplicate models and use them for its own training, Reuters reportedin February.
The memo andfollow-up cable, released just weeks before U.S. PresidentDonald Trump is set to visit Chinese President Xi Jinping in Beijing,promise to raise tensions in a long-running tech war between therival superpowers, which had been lowered by a detente brokered lastOctober.
