One of the leading companies in the field of artificial intelligence (AI) applications in the United States, Palantir, issued a warning on Wednesday, July 15th, through its Chief Technology Officer Shyam Sankar. Sankar warned that the practice of Chinese companies stealing AI technology achievements from Silicon Valley in the United States to develop their own next-generation AI models could have a substantial impact on the U.S. economy.
Sankar made these remarks while attending a defense innovation-related conference in Pennsylvania and being interviewed by Bloomberg Television. He stated that the core technology of the AI open-source models developed by China in recent years is essentially obtained through a method known as “distillation attack,” with much of the content originating from intellectual property stolen from cutting-edge laboratories in the United States.
A “distillation attack” refers to malicious entities utilizing numerous fake accounts to send a massive amount of queries to existing top AI models, such as the Claude model developed by the U.S. AI company Anthropic, and eavesdropping on the model’s response. Subsequently, they use this high-quality output data to train and fine-tune their own copied AI models, thus “freeloading” on the research and development achievements of their competitors, significantly reducing development costs.
Sankar urged the U.S. AI industry to enhance its own intellectual property protection mechanisms, emphasizing that this action is crucial for the economic interests of companies. His warning echoes the shared concerns of several top AI companies in the United States.
Just last month, the U.S. AI company Anthropic publicly accused the Chinese technology giant Alibaba Group of systematically conducting “distillation” operations against its Claude model through a large number of fake accounts.
Prior to this, Chinese artificial intelligence startups such as DeepSeek and MiniMax were also accused of using similar methods to develop competitive chatbot products.
In addition to intellectual property issues, Sankar also pointed out that the U.S. AI field currently faces another neglected danger: the growing public resistance to AI development. This is particularly evident in the opposition from residents to the construction of data centers needed for artificial intelligence in various places.
He cited the recent decision by New York Governor Kathy Hochul to postpone the approval process for new data centers by one year as an example. Sankar described that if the United States were to slow down its pace of AI development due to social pressure, the consequences would be no less than the price paid by the U.S. for abandoning nuclear energy development in the 1970s.
Regarding the French government’s decision last month to terminate its cooperation with Palantir in intelligence data tools and instead support local providers, Sankar expressed indifference, considering it as France’s policy choice to support its own industry.
Previously, France and other European countries have actively promoted independent technology policies, attempting to reduce their reliance on advanced technological systems, including AI, from the United States.
Sankar also emphasized that the strength of Palantir’s technology products is sufficient to prove their value. He criticized certain overseas practices attempting to undermine the company’s business, stating that some malicious spreading of rumors and misinformation clearly crossed the line of fair business competition.