[Former OpenAI Researcher Claims China Has Mastered All Code from U.S. AI Giants, Mocked by Industry] According to monitoring by Beating, former OpenAI researcher Will Depue claimed that China has fully mastered all GitHub repositories, Slack chat records, and internal documents of OpenAI and Anthropic. He further asserted that he wouldn’t be surprised to see 'reasonably deniable' stolen architectures in China’s open-source models. However, as he failed to provide any concrete evidence, Depue argued that his conclusions were primarily based on conversations with security researchers, accusing the public of overestimating the defenses of cutting-edge labs. Known for frequently posting outrageous statements on social media, Depue once claimed on April Fool’s Day that he would be joining DeepSeek. His remarks quickly became a laughingstock within the AI community. AI scholar Nathan Lambert criticized cutting-edge labs for repeatedly making self-serving nonsense statements, pointing out that the flow of technology fundamentally occurs through Silicon Valley’s talent mobility, bar meetups, and the lack of non-compete agreements. He emphasized that the true core advantages lie in computing power and engineering resources, not Slack chat records. AI researcher Elie Bakouch sarcastically responded in the exact same format, stating that without a doubt, all cutting-edge labs today have also fully mastered the technical reports and models of DeepSeek, Kimi, GLM, and MiniMax, mocking U.S. labs for constantly referencing China’s open-source achievements for free while simultaneously indulging in persecution fantasies. [Original Link]