研究人员推出了TRAP，这是一个旨在评估AI代理在完成任务的同时抵抗隐私提取能力的新基准。该基准评估了任务准确性和数据泄露之间的权衡，发现当前专有和开源模型都存在显著的隐私泄露。现有的基于提示的防御措施只能提供部分解决方案，并且常常以牺牲任务性能为代价。一种名为结构化私有字段隔离的新方法在不影响准确性的情况下防止泄露方面显示出潜力。 AI
影响 强调了在处理敏感数据的AI代理中实施强大隐私措施的关键需求，可能影响未来的模型开发和部署策略。
排序理由 该集群包含两篇介绍AI系统隐私审计新基准或方法的学术论文。
AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →
arXiv:2606.18996v1 Announce Type: cross Abstract: Agents are increasingly deployed in document-intensive workflows where sensitive private information is not an edge case but a routine input, e.g., an agent booking a flight needs passport numbers. In such settings, the agent must…
Agents are increasingly deployed in document-intensive workflows where sensitive private information is not an edge case but a routine input, e.g., an agent booking a flight needs passport numbers. In such settings, the agent must use private information to complete tasks accurat…
arXiv:2606.12733v2 Announce Type: replace Abstract: Privacy auditing provides an important safeguard by estimating the actual information leaked by a model, thus ensuring that theoretical privacy guarantees hold in practice. We study empirical privacy auditing for differentially …
Our society has become increasingly and dangerously comfortable sharing personal information over the last three decades.