US intelligence agencies (NSA, CISA, FBI) have accused six Chinese AI firms—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI—of industrial-scale distillation attacks against US frontier AI models since late 2024. The firms allegedly exploited AI inference APIs using fraudulent bulk accounts to extract capabilities from models like Claude, GPT, Gemini, and Grok, potentially saving billions in R&D costs.
Background
AI model distillation refers to extracting knowledge from large proprietary models to train smaller or comparable models at lower cost. The US-China AI competition has intensified concerns over intellectual property protection and model security.
- Source
- Ars Technica
- Published
- Sep 10, 2026 at 04:06 AM
- Score
- 7.0 / 10