25k fake accounts ran millions of queries to copy the model. Distillation attack exposed.
Anthropic caught Alibaba using 25,000 fraudulent accounts.
28.8 million conversations extracted Claude capabilities.
Attack ran from April 22 to June 5 2026.
Method called distillation. Train your model on answers from rival model.
DeepSeek Moonshot AI and MiniMax also involved.
Anthropic blocked Claude in China but it got bypassed.
Company sent urgent letter to Senate and White House asking for action.
🚨 CHINA LE ROBÓ LA IA A ANTHROPIC.
Y lo hicieron durante 45 días seguidos.
Anthropic acaba de acusar a Alibaba de lanzar el mayor ataque de robo de IA de la historia.
El objetivo: extraer las capacidades de Claude… sin pagar ni un dólar.
Cómo lo hicieron:
→ Crearon casi… https://t.co/KgB1aooXTC pic.twitter.com/ggisWbjFyw
— SONIA (@S0N_IA) June 24, 2026
Reuters full report https://www.reuters.com/world/china/anthropic-says-alibaba-illicitly-extracted-claude-ai-model-capabilities-2026-06-24/
CNBC on Anthropic letter https://www.cnbc.com/2026/06/24/anthropic-alibaba-distillation-campaign.html
