1 paper
Xinyue Huang, Xiaochun Cao, Wenyuan Yang
As LLMs become increasingly woven into everyday workflows, user queries sent to cloud hosted LLMs routinely mix task-essential content with task non-essential sensitive disclosures…