AI 导读
Anthropic 称已阻止未经授权使用 Claude 训练竞品模型的行为，指控阿里在 5 月至 7 月产生 151M+ 次交互，Moonshot 为 23M+ 次。它还称 DeepSeek 在 14 天内进行了 12M+ 次蒸馏攻击，并指 Moonshot 与 DeepSeek 在用户不知情下将客户请求转发给 Claude，部分含敏感数据。三家公司未立即回应 CNBC。
推荐理由
Anthropic 指控阿里、Moonshot 与 DeepSeek 未授权用 Claude 训练竞品模型，并给出各家调用量数字，可用于观察模型蒸馏争议的走向。
正文
Anthropic says it disrupted unauthorized efforts to train rival models using Claude, alleging Alibaba generated 151M+ exchanges and Moonshot 23M+ between May and July. It attributed 12M+ distillation attacks to DeepSeek over 14 days.
It also alleges Moonshot and DeepSeek forwarded customer requests to Claude without users’ knowledge, with some containing sensitive data.
The three companies did not immediately respond to CNBC.
来源：@thexpin · x.com