started · updated
Anthropic Accuses Alibaba of Large‑Scale AI Model Theft
Anthropic, a U.S. AI developer, alleges that operators linked to China’s Alibaba Group used roughly 25,000 fraudulent accounts to conduct about 28.8 million interactions with its Claude model between late April and early June. The company says the campaign targeted Claude’s most advanced capabilities, such as agentic reasoning and software‑engineering functions, in an “industrial‑scale” distillation attack that copies U.S. AI research without bearing the development costs.
Anthropic sent a letter on 10 June to U.S. senators Tim Scott and Elizabeth Warren and to White House officials, urging Congress to impose penalties on firms that engage in such illicit extraction, to tighten export‑control rules on AI chips, and to facilitate information‑sharing among AI companies. The firm also cited Department of Defense reports that have linked Alibaba to China’s military.
Alibaba has not commented on the specific allegations. Following the report, Alibaba’s shares fell 3‑5 % on both the New York and Hong Kong exchanges. The dispute adds to rising U.S.–China tensions over artificial‑intelligence leadership and follows earlier Anthropic warnings about similar campaigns by other Chinese labs such as DeepSeek and MiniMax.