Summary

Anthropic claims that three Chinese tech labs (Deep Seek, Moonshot, and MiniMax) are conducting distillation attacks to extract AI model capabilities, allegedly using over 24,000 fraudulent accounts to generate 16 million exchanges. The accusation suggests these labs aim to illicitly transfer American AI model capabilities into Chinese military, intelligence, and surveillance systems. While the speaker acknowledges the seriousness of the claim, they also express skepticism about its full legitimacy and potential broader implications for the tech industry.