Former OpenAI researcher Will Depue claims China has already obtained every GitHub repository, Slack conversation, and internal document belonging to OpenAI and Anthropic. He even said he wouldn't be surprised to see "plausibly deniable" stolen architecture in Chinese open-source models.
https://twitter.com/willdepue/status/2068118253633737077
Despite offering zero evidence, Depue argued his conclusions came from conversations with security researchers and accused the industry of overestimating frontier labs' defenses. Depue, who frequently posts outlandish claims on social media, also announced on April Fools' Day that he was joining DeepSeek.
The AI community quickly turned his claims into a laughingstock. AI scholar Nathan Lambert called it "self-serving nonsense" from frontier labs, pointing out that technology naturally spreads through Silicon Valley's talent mobility, bar gatherings, and lack of non-compete agreements — the real advantages are compute and engineering resources, not Slack logs.
AI researcher Elie Bakouch responded with a perfectly mirrored sarcastic post: "Without a doubt, today all frontier labs also fully possess the technical reports and models of DeepSeek, Kimi, GLM, and MiniMax," skewering how US labs freely learn from Chinese open-source contributions while wallowing in persecution fantasies.