Former OpenAI researcher claims China has mastered all the code of American AI giants, mocked by industry peers

According to Beating monitoring reports, former OpenAI researcher Will Depue claimed that China today has fully mastered all of OpenAI and Anthropic’s GitHub code repositories, Slack chat records, and internal documents. He also vowed that he would not be surprised if, in China’s open-source models, he came across “reasonably deniable” stolen architectures.

Because he could not produce any solid proof, Depue argued that these conclusions mainly came from his communications with fellow security researchers, accusing the outside world of overestimating the defenses of cutting-edge labs. As a blogger who frequently posts outrageous statements on social platforms, Depue also posted on April Fool’s Day claiming that he would be joining DeepSeek.

These remarks quickly became a laughingstock in the AI community. AI scholar Nathan Lambert criticized cutting-edge labs for constantly publishing self-serving nonsense and noted that the circulation of technology is essentially achieved through the flow of talent in Silicon Valley, bar gatherings, and the lack of non-compete restrictions—while the real core advantages are computing power and engineering resources, not the chat logs in Slack.

AI researcher Elie Bakouch, using the exact same format, replied with a biting parody: without a doubt, today all cutting-edge labs have also fully grasped the technical reports and models of DeepSeek, Kimi, GLM, and MiniMax—mocking American cutting-edge labs for referencing the results of Chinese open source for free every day while spending all day trapped in forced persecution fantasies.