In brief
Anthropic, a leading U.S. AI developer, claims that Alibaba-linked operators performed a large‑scale distillation attack on its Claude model, using many fake accounts to harvest advanced capabilities. The letter sent to Senators Tim Scott and Elizabeth Warren urges Congress to penalize entities that steal AI innovations and strengthen protective measures. Alibaba has denied the accusations and recently sued the U.S. government to remove its name from a Pentagon blacklist, while Washington’s defense department has linked Alibaba to the military. The dispute highlights a growing tension over AI technology theft amid heightened competition between U.S. and Chinese firms.
US artificial‑intelligence firm Anthropic has filed a formal complaint against Chinese e‑commerce giant Alibaba, claiming that Alibaba’s operators carried out a "distillation attack" against its Claude model. The attack allegedly involved close to 29 million queries sent to Claude through thousands of fraudulent accounts, which the company says allowed Alibaba‑linked parties to copy sophisticated functionalities such as long‑form task handling and decision‑making strategies.
The letter, addressed to Senators Tim Scott and Elizabeth Warren on 10 June, urges Congress to impose penalties on firms that engage in this type of technology theft and to bolster safeguards against large‑scale extraction of AI capabilities. It also warns that similar attacks could undermine the United States’ military competitiveness, citing U.S. Department of Defense concerns about ties between Alibaba and the Chinese armed forces.
Alibaba has denied the allegations and has recently filed a lawsuit to remove its name from a Pentagon blacklist that designates companies under potential military scrutiny. The company’s legal move comes amid a broader U.S. effort to curb the export of advanced AI systems to China.
Anthropic’s complaint follows a pattern of accusations by U.S. AI developers against Chinese competitors. OpenAI, for instance, has previously alleged that Chinese firms were exploiting distillation methods to train rival models at a fraction of the R&D cost. These disputes underscore the growing global rivalry over ownership of AI intellectual property and the potential security repercussions of stolen data.
Photo: Dario Amodei, CEO of Anthropic, in a corporate interview setting. Source: Bloomberg via Getty Images.
