Anthropic 指控阿里巴巴旗下 Qwen AI 实验室相关的运营商，使用近 2.5 万个虚假帐户与其 Claude AI 模型进行了高达 2880 万次对话，以非法获取其软体工程和代理推理等关键能力。

此类被称为「对抗性蒸馏」的行为引起了美国开发商与国会的高度警惕，参众两院正推动立法，计划对不当获取美国 AI 模型数据来训练竞争产品的中国企业实施黑名单或制裁。

尽管 Anthropic 呼吁美政府与行业加强合作以应对此类蒸馏攻击，但该公司目前正因安全疑虑面临川普政府的出口管制，已暂停其 Fable 5 和 Mythos 5 等部分先进模型的服务。

Anthropic accused operators linked to Alibaba's Qwen AI lab of using nearly 25,000 fraudulent accounts to conduct 28.8 million exchanges with Claude to illicitly access key capabilities like software engineering and agentic reasoning. (Key numbers: 2.5, 2880)

This practice of adversarial distillation has alarmed US developers and Congress, prompting bipartisan legislative efforts to blacklist or sanction Chinese firms that improperly access US AI outputs to train competing models.

While Anthropic urged stronger government-industry cooperation to combat these distillation attacks, the company itself is embroiled in an export control dispute with the Trump administration and has suspended services for some of its advanced models. (Key numbers: 5)