According to CoinMeta, Anthropic has written to the U.S. Congress, accusing Alibaba of illegally obtaining the intellectual property rights of its Claude model through large-scale distillation and attack. Alibaba is alleged to have registered 25,000 fake accounts and initiated 28.8 million interactions within three months, extracting 57.6 billion token data points to train its own large language models. These actions were intended to mimic normal user behavior in order to evade detection.