Chinese AI labs are training models on Nvidia's Blackwell despite export controls, says ex-OpenAI engineer
Forensic report finds Blackwell's fingerprint in DeepSeek, Moonshot AI, MiniMax and Ant Group models
Former OpenAI engineer Will DePue, who recently launched an AI research consultancy, has uncovered evidence that Chinese AI labs are training models on Nvidia's restricted Blackwell GPUs, despite export controls. This surprising discovery was made by analyzing the hidden signatures of GPU training in published AI models. DePue found that models from DeepSeek, Moonshot AI, MiniMax, and Ant Group were indeed initialized on Nvidia Blackwell systems.
These findings, published by DePue's consultancy, suggest that the current chip export controls are ineffective, as Chinese labs have found a way to circumvent them. The report highlights the need for stricter regulations to prevent Chinese AI labs from accessing cutting-edge AI technology.
Written by urgent.news from The Register Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
- Chinese AI labs are training models on Nvidia's Blackwell despite export controls, says ex-OpenAI engineer theregister.com
- Export-banned Nvidia chips are behind China’s most powerful AI models—uh oh, Jensen! techcrunch.com
- Documents tie banned Nvidia AI chips to Alice Huang-linked Shanghai data center qz.com