Home›Crypto›Market Structure›Moonshot accused of using American AI model outputs to…

Moonshot accused of using American AI model outputs to train Kimi

U.S. agencies say the alleged practice used billions of tokens from models including Claude, GPT, Gemini and Grok across millions of requests, via proxies and multiple accounts.

Reported · Written up by our desk · Market Structure

U.S. security agencies, including the NSA, FBI and CISA, accused six Chinese artificial intelligence companies of extracting capabilities from leading American frontier models at industrial scale, using the outputs to train their own systems, according to CoinDesk.

The advisory said the companies used billions of tokens across millions of requests from models such as Claude, GPT, Gemini and Grok, often relying on proxy services and multiple accounts to avoid detection and usage limits. It described the approach as malicious knowledge distillation, and said it likely took place with the Chinese government's awareness.

The allegations include Moonshot AI, maker of Kimi, whose strong coding performance had previously helped rattle markets. The agencies said Moonshot used outputs from Claude and GPT to improve Kimi K3, an open-weight model that topped a widely watched coding benchmark in July and briefly triggered a broader AI selloff.

CoinDesk noted that Chinese AI stocks were hit, with Z.ai falling about 27% and MiniMax down about 16%, while semiconductor shares weakened across Asia and bitcoin and other major cryptocurrencies fell. The pressure continued into the next session, with South Korea’s Kospi dropping another 3.5%.

Latest closeBitcoin $79,100.85 ▼0.0%|Kospi 6,995.39 ▲4.6%

More like this

Sources