ServiceNow

Malice in Agentland: Down the Rabbit Hole of Backdoors in the AI Supply Chain

Agentland中的恶意行为：深入AI供应链后门问题

Léo Boisvert, Abhay Puri, Chandra Kiran Reddy Evuru, Nazanin Sepahvand, Nicolas Chapados, Quentin Cappart, Jason Stanley, Alexandre Lacoste, Krishnamurthy Dj Dvijotham, Alexandre Drouin

机构 * ServiceNow Research ； Mila - Qu\'ebec AI Institute 纠错

AI总结 研究探讨了在交互数据上微调AI代理时引入的安全漏洞，提出三种供应链威胁模型，证明通过污染少量演示即可使代理泄露用户信息。

Comments 27 pages