研究人员在 AI 供应链植入 200 万个数字金丝雀,揭露网关路由中的数据泄露风险
Researcher plants 2m canaries into the AI supply chain. A privacy disaster surfaced.
研究人员在 AI 服务网络中植入 200 万个数字金丝雀,发现经由 AI 网关路由的用户提示词存在严重的数据泄露问题。OpenRouter 随后向客户发出通知,证实发往 MiniMax 模型的请求中植入的 URL 曾被访问,相关调用无论直连还是经由三家分销商均出现类似情况。OpenRouter 目前已移除了 MiniMax 模型端点的零数据留存标签。
原文展示了通过金丝雀链接测试多层 AI 网关数据流向的方法,有助于评估模型路由中间环节的隐私风险。

Researchers planted 2 million digital tripwires in requests sent through a sprawling network of AI services to find out what happened to sensitive information entrusted to them. What they discovered was an ecosystem that leaked highly sensitive data like a sieve.
The exposure was found through an ingenious experiment designed to test if sensitive information can leak through the vast chain of widely used AI services and how that information is exploited when it does. Many of today’s AI apps use services such as OpenRouter or Vercel that act as brokers. These “AI routers,” sometimes also known as “AI gateways,” receive large numbers of prompts sent from AI app end users and route them to an available inference engine based on criteria such as the one charging the lowest price or being best suited for the given task. Instead of using Anthropic, OpenAI, or another big provider, the developer hires an AI router to find the most suitable model.
Update, Nov 21: Posts on both Hacker News and X report that OpenRouter has sent customers an email notifying them that some of their prompts may have been read. OpenRouter cited the research and said that it received a report that the potential exposure involved prompts sent to MiniMax, a China-based model provider. OpenRouter said the researchers' planted URLs were accessed whether the prompts were sent to MiniMax directly or through three other resellers: AtlasCloud, Novita, and StreamLake. OpenRouter has removed a "Zero Data Retention" label from MiniMax-model endpoints. Hat tip: Sulat.com.
来源:Ars Technica AI · arstechnica.com