Researcher finds Blackwell GPU signatures in Chinese AI models
Source published ·Modelwire updated
Original coverage: The Decoder ↗·How Modelwire adds context

The development
A former OpenAI researcher has published forensic evidence that Chinese AI labs including DeepSeek, Moonshot AI, MiniMax, and Ant Group are training models on Nvidia's Blackwell GPUs, circumventing US export restrictions. By analyzing characteristic signatures embedded in published model weights, the researcher identified at least five training runs representing over half a billion dollars in hardware investment. This finding exposes a critical gap between US export control enforcement and actual chip deployment in China's AI infrastructure, raising questions about the effectiveness of current sanctions and the pace of Chinese model development.
Modelwire’s AI-generated summary of coverage from The Decoder.
Modelwire analysis
Analyst takeOur AI-generated reading of the wider context and the next developments to watch.
The half-billion-dollar hardware investment figure in the summary is striking precisely because it mirrors the figure from a separate story two days earlier, suggesting these two data points may be measuring overlapping or related capital flows rather than independent violations. The researcher's methodology, fingerprinting GPU signatures from published model weights, is also a novel enforcement tool that no government body has publicly claimed to use.
This finding lands at the center of a cluster of stories Modelwire has tracked over the past week. The Verge's reporting on Nvidia H200 sales approved to a shell company in China (November 7) established that export controls have procedural gaps at the point of sale. The Blackwell fingerprinting evidence suggests those gaps extend well past the transaction itself, into active training runs. Meanwhile, the ChinaTalk investigation we covered on November 11 documented roughly half a billion dollars in RL training environments flowing to Chinese labs through data brokers, which means hardware access and training data access appear to be advancing in parallel, not sequentially.
Watch whether the Commerce Department's Bureau of Industry and Security publicly acknowledges the fingerprinting methodology within the next 60 days. If they adopt or cite it in an enforcement action, the technique becomes a durable part of the control regime. If they ignore it, that tells you something concrete about enforcement appetite.
This interpretation is generated from the summary above and available source metadata. Our methodology · Report an error
MentionsWill DePue · OpenAI · Nvidia Blackwell · DeepSeek · Moonshot AI · Ant Group
How this coverage is produced
Modelwire uses AI to generate summaries and context from source headlines, snippets, and selected archive coverage. Automated checks do not verify every claim, and items are not routinely reviewed by a person before publication. Zacaria Solis operates the site. Read the linked source for the full evidence and report errors through our corrections process.
Modelwire summarizes, we don’t republish. The Decoder originally reported this story as “Former OpenAI researcher says he found Nvidia Blackwell fingerprints hidden inside six Chinese AI models”. The full content lives on the-decoder.com. If you’re a publisher and want a different summarization policy for your work, see our takedown page.