Data theft attacks on Claude cut by 99% with new measures: Anthropic

Anthropic said its new security measures have reportedly shown a 90-99% reduction in data stolen through distillation attacks. It also identified nearly 200 million exchanges across five campaigns allegedly aimed at extracting its model's capabilities. This comes a month after Anthropic found that Chinese models, including DeepSeek and Moonshot, unlawfully used Claude to answer users' queries.