Tech Tensions: Moonshot AI Accused of Illicitly Replicating US AI Model

Chinese AI startup Moonshot AI is accused of illegally copying Anthropic's Fable model to create its K3 model, as stated by White House official Michael Kratsios. The U.S. emphasizes legitimate AI development but warns against covert tech theft, amid allegations of Moonshot AI's sophisticated evasion tactics.

In a charged statement, Michael Kratsios, director at the White House Office of Science and Technology Policy, accused Chinese artificial intelligence startup Moonshot AI of illegally reproducing Anthropic's Fable model to craft its K3 model. Kratsios aired these concerns on his official X account, delineating the fine line between acceptable AI distillation practices and unauthorized technological acquisitions.

Kratsios emphasized the United States' robust support for the transparent and fair advancement of AI technologies, noting that legitimate distillation efforts can enhance AI efficiency. However, he sternly denounced any large-scale, clandestine attempts to usurp proprietary U.S. technology, which undercuts American research endeavors. He underscored the critical role of lawful AI distillation in fostering innovation but condemned underhanded practices that aimed to misappropriate U.S. advancements.

According to Kratsios, Moonshot AI's replication drive involved sophisticated internal schemes to perform extensive distillation of American AI models, allowing stealthy switching between access methods to evade detection. Additionally, claims were made regarding Moonshot AI's acquisition of GB300-equipped servers in Thailand, potentially for training purposes, adding another layer to the controversy enveloping the startup's activities.