IC Technology & Cyber — 2026-09-09

NSA CISA and FBI Joint Advisory Names Six Chinese AI Firms for Industrial-Scale Distillation of US Frontier Models

BLUFBy elevating distillation to a formal national-security designation, the advisory almost certainly accelerates US provider access restrictions within days and will likely draw a coordinated Chinese rebuttal within 30 days.

The NSA, CISA, and FBI on September 8 named six China-based AI companies, DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI, in a joint advisory alleging industrial-scale distillation campaigns against U.S. frontier models since at least late 2024 1. The agencies said the firms extracted billions of tokens across millions of exchanges from Claude, GPT, Gemini, and Grok variants, "likely with Chinese government awareness," and characterized the activity as forming "the core, not merely a supplement" of the firms' AI development strategy 12. The advisory attributes specific campaigns to individual firms: between late 2024 and mid-2025 DeepSeek allegedly distilled Claude 3.7, Sonnet 4, Sonnet 4.5, and Opus 4.1 alongside Gemini 2.5 Pro/Flash Preview, GPT-4, GPT-4o, GPT-4 Mini, GPT-4 Nano, GPT-5, and Grok 4 to train its R1 and V3 systems, while Moonshot AI allegedly extracted Claude Fable 5 data for Kimi-K3 and GPT-4o data for Kimi-K2 13. The agencies described tactics including fraudulent accounts, bulk subscription procurement, and "transfer station" proxies that obfuscate user metadata and evade geographic restrictions, and recommended U.S. AI firms subtly alter outputs for suspected distillers without notifying them 14. None of the six named companies had issued a public response at the time of publication 4.

Analysis
This advisory shifts distillation from a bilateral IP dispute into a formal national-security designation, giving US labs cover to deploy the recommended countermeasure of covertly degrading outputs for suspected Chinese accounts without disclosure, an expansion from prior individual accusations by Anthropic and the White House OSTP into a joint six-firm interagency designation. Named firms almost certainly face tightened API access and subscription vetting from US providers within days. The specificity on model versions and token volumes may instead reflect data agencies obtained through cooperating US AI firms' own abuse-detection systems, making this closer to an official endorsement of industry claims than independent government collection. A coordinated Chinese government response or firm-level denial is likely within the next 30 days given Beijing's pattern of measured pushback on prior distillation claims; absent that, providers proceed unchallenged to implement silent output alteration. Confidence is moderate: the technical claims rest on a single primary source whose attribution methodology is not independently verifiable in open reporting.
4 sources
  1. China-Based Artificial Intelligence Companies Conducting Industrial-Scale Distillation Campaigns Against U.S. AI Companies - CISA
  2. Intelligence agencies warn of China's large-scale AI model distillation efforts - Nextgov/FCW
  3. Feds accuse China of 'systematic' distillation of U.S. AI models - CyberScoop
  4. US Names Six Chinese AI Firms Accused of Stealing Claude, GPT and Gemini Capabilities - International Business Times

View in full brief →

UNCLASSIFIED // OPEN SOURCE