Chinese AI labs are training models on Nvidia's Blackwell despite export controls, says ex-OpenAI engineer

Comparison summary

AI-assisted · Cerebras / Llama · Nov 13, 2026, 4:27 PM · inspect sources below rather than trusting this alone

The Register reported that former OpenAI engineers claim Chinese AI laboratories are circumventing US export controls to train models on Nvidia’s Blackwell chips. The allegation suggests that despite regulatory restrictions, advanced hardware is reaching Chinese developers for large-scale model training. This claim originates from a single source, an ex-OpenAI engineer, and has not been independently verified by Nvidia or Chinese authorities.

Coverage divergence is minimal because both entries in this cluster are identical, with the second source merely republishing The Register’s article. Consequently, no contrasting framing exists between the two provided texts; both present the allegation as a factual report without additional editorial context. The lack of distinct outlets prevents a comparison of how different political or geographic perspectives might interpret the breach of export controls.

The cluster lacks independent verification from Nvidia, the US Department of Commerce, or Chinese state media, leaving the core claim unconfirmed. Additionally, no outlet provided technical details on the specific circumvention methods or the scale of the alleged hardware diversion. This absence of corroborating evidence represents a significant blind spot, as the story relies entirely on a single whistleblower’s assertion without broader industry or governmental confirmation.