China‑Linked AI Theft: Alibaba Accused of Stealing Anthropic’s Claude Model
US AI powerhouse Anthropic has publicly accused Chinese e‑commerce titan Alibaba of conducting an “illicit” distillation attack on its Claude model. The letter, sent to Senators Tim Scott and Elizabeth Warren, argued that Alibaba manipulators ran close to 29 million exchanges with Claude through thousands of fraudulent accounts, systematically harvesting the model’s most potent capabilities.
In the world of machine‑learning, a distillation attack is a covert method of copying a high‑performance model’s answers to train an inferior one. Anthropic warned that such tactics allow rival firms to reap U.S. research dividends without the steep cost normally required to build cutting‑edge AI, effectively turning billions of dollars of American investment into a free subsidy for geopolitical competitors.
Anthropic cited concerns that the extracted knowledge could be weaponised, stating that the campaign poses a threat that could even extend to U.S. military capabilities. It highlighted that other Chinese enterprises—including BYD and Baidu—have similar ties to the Chinese military, according to the Department of Defense.
The allegations are part of a growing backlash against Chinese AI firms, with several U.S. developers previously claimed similar covert copying practices. Alibaba, which has denied the accusations, recently filed a lawsuit against the U.S. government seeking removal from the Pentagon black‑list, a move that underscores the transactional tension between tech sovereignty and international policy.
While this dispute centers on commercial power, the implications stretch into climate science. AI models like Claude underpin advanced climate‑prediction tools and resource‑efficient technologies—fields where data integrity and sovereign control are paramount. The risk of illicit exfiltration threatens not only national security but the timely, reliable development of AI solutions that address climate change, demanding a reevaluation of safeguards in advanced modelling platforms.
Anthropic’s accusation is a crystallisation of the growing AI arms race, with how nations protect core capabilities becoming as critical as bullets in a conflict. The company’s call for congressional oversight and tighter security measures reflects a broader push to fortify U.S. innovation against foreign harvesting that could destabilise ethically guided technology development, including climate‑oriented AI systems.
