Washington accuses Chinese startup of stealing Anthropic AI
Tap title to read story
Washington officials have accused Beijing-based Moonshot AI of large-scale intellectual property theft from U.S. artificial intelligence developer Anthropic, according to a formal announcement by senior White House technology staff. Michael Kratsios, Director of the White House Office of Science and Technology Policy, said federal authorities have specific information that Moonshot AI systematically distilled Anthropic’s frontier model, Fable 5. U.S. officials allege Moonshot AI built an internal system to send massive automated queries to American models, rotating access methods to bypass security filters and evade detection. Outputs were allegedly harvested as training data for Moonshot’s Kimi K3 model, unveiled in July 2026.
More context
Timeline
- July 2026 Kimi K3 model publicly unveiled
- Earlier this year Moonshot deploys internal query platform
- Subsequent months automated queries target U.S. models
- Following that security filters allegedly circumvented repeatedly
- Recently White House gathers specific technical evidence
- This week Washington formally announces theft accusations
- Today officials warn of threatened research leadership
Why it matters
This is about your MONEY and PRIVACY. If Moonshot AI did steal from Anthropic, it could mean cheaper AI tech. But it also means your data might be less secure. Check what AI services you use. Are they safe?
JQJO Editorial Desk · Human reviewed · AI assisted. Report a correction.
