Written edition
Beijing probes DeepSeek and Moonshot for secretly routing queries to Anthropic
A quiet Wednesday carries one loud undercurrent: AI's back-end plumbing keeps getting exposed, from Chinese labs quietly leaning on a rival's model to a tier-one automaker ceding its own autonomy stack to a startup.
Elsewhere, speed and scale are the day's currencies, in payments, voice AI, and blockchain settlement alike.
Beijing investigates Moonshot and DeepSeek for routing queries to Anthropic
Chinese regulators are investigating Moonshot AI and DeepSeek, the country's two most prominent AI labs, over allegations they secretly routed some user queries to Anthropic's Claude model and served the answers back as their own, without telling users. The probe is examining whether this broke data-export or AI-security rules, and it lands awkwardly for Moonshot as it prepares to go public. Beyond the regulatory risk, it raises a harder question for the entire Chinese AI industry: how much of the frontier capability on display has been genuinely home-built versus quietly borrowed from foreign models to compete on launch timelines. That distinction matters for anyone assessing China's real AI independence.
