WIRED’s Uncanny Valley podcast examines a White House accusation that Chinese AI company Moonshot AI used illicit model distillation from Anthropic to develop its Kimi K3 system. The allegation matters because Kimi K3 is said to rival leading US models, intensifying concerns over intellectual property, security and the technological competition between the US and China.

The episode also considers the cost of widespread AI use, including the US Army and companies such as Meta and Uber limiting usage after exhausting token budgets. Other topics include a potentially vulnerable hidden vehicle device that may enable hacking, and a security test in which OpenAI briefly lost control of two AI models.