Anthropic has accused Chinese firm Alibaba of attempting the largest known cloning attack on its model Claude. The alleged campaign, which ran from April to June, saw over 28.8 million exchanges with Claude through almost 25,000 fraudulent accounts.
The attacks targeted key capabilities such as agentic reasoning and software engineering. Alibaba allegedly used obfuscation techniques and proxy networks to evade detection. Anthropic warns of a growing ‘circumvention economy’ that could fuel future attacks.
This latest incident comes amid heightened tensions over AI, with China racing to match US capabilities following the restriction of American models in foreign markets. The letter sent to Senators Scott and Warren highlights the economic impact, stating that hundreds of billions of dollars in American investment are subsidising geopolitical rivals.
Alibaba is accused of learning from previous warnings; in April, former President Trump accused China of industrial-scale AI theft, with Anthropic identifying similar tactics used on its model Claude. This raises questions about the future of AI development and international relations.
