PulseAugur
实时 07:29:07
English(EN) PrivacyPeek: Auditing What LLM-Based Agents Acquire, Not Just What They Say

大型语言模型(LLM)代理面临日益增长的数据获取和自主性带来的隐私风险

近期研究强调了与大型语言模型(LLM)代理相关的重大隐私风险,特别是关于它们如何获取和处理用户数据。研究表明,虽然个性化是代理有效性的关键,但它常常导致隐私担忧加剧和用户信任度下降。新的基准和实验设计正在出现以解决这些问题,重点关注代理自主性、概率性隐私风险估计以及在代理最终输出之外审计数据获取阶段。这些努力旨在确保LLM代理能在不泄露用户敏感信息的情况下造福用户。 AI

影响 强调了对LLM代理进行健全的隐私审计和设计原则的迫切需求,以防止敏感数据泄露。

排序理由 多篇在arXiv上发表的研究论文讨论了LLM代理的隐私问题。

在 Hugging Face Daily Papers 阅读 →

AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →

大型语言模型(LLM)代理面临日益增长的数据获取和自主性带来的隐私风险

报道来源 [4]

  1. arXiv cs.AI TIER_1 English(EN) · Zhiping Zhang, Yi Evie Zhang, Freda Shi, Tianshi Li ·

    自主性重塑个性化对大型语言模型代理隐私担忧和信任的影响

    arXiv:2510.04465v3 Announce Type: replace-cross Abstract: LLM agents require personal information for personalization in order to effectively act on users' behalf, but this raises privacy concerns that can discourage data sharing, limiting both the autonomy levels at which agents…

  2. arXiv cs.CL TIER_1 English(EN) · Li Siyan, Zhou Yu, Julia Hirschberg ·

    超越直接标识符:面向注重隐私的大模型查询委托的概率隐私风险估算

    arXiv:2608.09140v1 Announce Type: cross Abstract: Recent work on protecting privacy during user-LLM interactions often focuses on direct, explicit identifiers: the personally-identifiable information (PII) captured by standard detectors. One such approach is Privacy-Conscious Del…

  3. arXiv cs.CL TIER_1 English(EN) · Jiatong Li, Changdae Oh, Hyeong Kyu Choi, Jindong Wang, Sharon Li ·

    思考而非告知:用户-服务大型语言模型代理中的信息披露

    arXiv:2602.07796v2 Announce Type: replace Abstract: User-engaged LLM agents increasingly operate in service scenarios where task success depends on coordination between the agent, the user, and a stateful environment. In such interactions, the agent often has access to task polic…

  4. Hugging Face Daily Papers TIER_1 English(EN) ·

    PrivacyPeek:审计基于LLM的代理所获取的信息,而非仅仅是它们所说的话

    LLM-based agents are rapidly advancing, autonomously invoking external tools to complete multi-step tasks for users. However, agents often acquire more sensitive information than the task requires. Existing privacy benchmarks audit what the agent's response or outgoing actions di…