US agencies accused six Chinese AI companies of stealing US model capabilities through large-scale API abuse. The FBI, NSA, and CISA claimed Deep Seek, Moonshot, Alibaba, Mini Max, Step Fun, and Z. AI used fake accounts and prompt injection to extract hidden reasoning from models like Claude, Gemini, and GPT.

The agencies urged US firms to improve detection, monitor suspicious usage patterns, and consider altering responses to reduce theft. They also advised strengthening identity checks and tracking enterprise subscriptions, while noting potential privacy trade-offs.