The NSA, CISA, and FBI issue a joint advisory warning that Chinese AI companies, including DeepSeek, are conducting "industrial-scale" distillation campaigns (Reuters)
ThinkingNews Desk · how this was written
U.S. intelligence agencies—the NSA, FBI and CISA—issued a joint advisory warning that six Chinese artificial-intelligence companies are conducting industrial-scale model-distillation campaigns that extract capabilities from American frontier AI models. The firms—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI—have reportedly harvested billions of tokens from Anthropic’s Claude, OpenAI’s GPT, Google’s Gemini and xAI’s Grok to train their own systems.
Written from all 3 reports below, not from any single one.
How it was reported
- TechMeme·The NSA, CISA, and FBI issue a joint advisory warning that Chinese AI companies, including DeepSeek, are conducting "industrial-scale" distillation campaigns (Reuters)
The NSA, CISA and FBI issued a joint advisory warning that Chinese artificial-intelligence companies, including DeepSeek, are running industrial-scale model-distillation campaigns that could expose U.S. data. The agencies say the operations involve large-batch extraction of proprietary model weights, posing a significant national-security risk.
- The Next Web·US intelligence advisory names six Chinese AI firms and lists the US models each one targeted
Three U.S. agencies—the NSA, FBI and CISA—identified six Chinese artificial-intelligence companies that have been systematically extracting capabilities from American frontier models beginning in late 2024. Their joint advisory details the specific U.S. models each firm pursued and describes the disguises used to request the data.
- Engadget·US authorities accuse Chinese AI companies of industrial-scale campaigns to copy American models
U.S. agencies report that Chinese firms DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI have extracted billions of tokens from Anthropic’s Claude, OpenAI’s GPT, Google’s Gemini and xAI’s Grok since 2024 to train their own models. DeepSeek used Claude, Gemini, GPT and Grok data for its R1 model, while Moonshot AI incorporated Claude’s Fable and GPT-4o data into Kimi K3 and Kimi K2, respectively, and the advisory includes mitigation steps for U.S. companies.
- The Next Web·China rejects the US distillation advisory as unfounded accusations and smears
China’s foreign ministry dismissed a joint U.S. intelligence advisory that alleged Chinese developers were extracting capabilities from American frontier AI models at industrial scale, labeling the claims “unfounded accusations and smears.” Spokesperson Mao Ning asserted that China’s AI progress results from high-level technological self-reliance and national strength.
Related stories
- Diplomatic cable: US State Department has ordered a global push to bring attention to what it says are efforts by Chinese companies to steal IP from US AI labs (Raphael Satter/Reuters)3 outlets
- China’s AI models have Trump’s AI world at war with itself3 outlets
- The White House Is Trying to Figure Out What to Do About Chinese AI4 outlets
- Beijing hits back at Anthropic CEO's call to curb China's AI development3 outlets