Anthropic Accuses Alibaba of Illicitly Extracting Claude AI Capabilities

US artificial intelligence company Anthropic has leveled serious accusations against Chinese technology and e-commerce behemoth Alibaba, alleging a systematic and large-scale effort to illicitly extract the core capabilities of its Claude AI model. According to a letter sent by Anthropic to the US Congress, and reported by The Financial Times, Alibaba mobilized around 25,000 fake accounts to engage in over 28.8 million interactions with Claude over a period spanning from April 22 to June 5. Anthropic characterized this operation as a 'distillation' effort, a technique where a less advanced AI model is trained on the outputs generated by a more sophisticated one. The company highlighted that Alibaba specifically targeted areas where Claude excels, such as complex software engineering tasks, advanced agent reasoning, and the execution of long-term projects. This suggests a deliberate attempt to replicate Claude's advanced functionalities within Alibaba's own AI offerings. The AI startup stated that despite its policies of not providing Claude services to Chinese companies, Alibaba circumvented these restrictions through the use of fraudulent accounts. Anthropic views these actions as a significant breach of its terms of service and a calculated move to gain an unfair advantage. The scale of the operation, involving millions of exchanges, underscores the industrial nature of the alleged intellectual property theft. These allegations emerge amidst heightened tensions regarding intellectual property and AI development between the US and China. Anthropic's letter to the US Senate Financial Services Commission on June 10 detailed the extent of the campaign, emphasizing that it was not merely typical chatbot usage but a concerted effort to collect Claude's outputs in bulk to enhance Alibaba's proprietary AI models. The incident highlights ongoing concerns within the AI industry about model security and the protection of advanced AI capabilities.