Anthropic flags Alibaba's 'brazen' distillation of Claude

anthropic has accused chinese tech giant alibaba of 'illicitly' extracting data from its claude models to train competitors, calling it the 'largest known' distillation attack. the firm alleges alibaba used thousands of poe accounts to harvest text outputs, bypassing safeguards designed to prevent unauthorized model replication. this incident underscores the escalating tension between western ai labs and chinese competitors seeking to close the capability gap through reverse-engineering. the revelation comes as geopolitical pressure increases to secure ai supply chains against foreign adversaries.