mlsanityData to 7 Oct 2026

Say Something Else: Rethinking Contextual Privacy as Information Sufficiency

arXiv 2604.06409 ↗First posted 7 Apr 2026cs.CR cs.AI cs.CLcited by 0

Not on the map. Why?

Abstract (arXiv)

LLM agents increasingly draft messages on behalf of users, yet users routinely overshare sensitive information and disagree on what counts as private. Existing systems support only suppression (omitting sensitive information) and generalization (replacing information with an abstraction), and are typically evaluated on single isolated messages, leaving both the strategy space and evaluation setting incomplete. We formalize privacy-preserving LLM communication as an Information Sufficiency (IS) task, introduce free-text pseudonymization as a third strategy that replaces sensitive attributes with functionally equivalent alternatives, and propose a conversational evaluation protocol that assesses strategies under realistic multi-turn follow-up pressure. Across 792 scenarios spanning three power-relation types (institutional, peer, intimate) and three sensitivity categories (discrimination risk, social cost, boundary), we evaluate seven frontier LLMs on privacy at two granularities, covertness, and utility. Pseudonymization yields the strongest privacytextendash utility tradeoff overall, and single-message evaluation systematically underestimates leakage, with generalization losing up to 16.3 percentage points of privacy under follow-up.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 15 of its 27 references that are in the map. First posted on arXiv 7 Apr 2026.

Builds on

15 of 27 references.

References · 15

  1. 2023Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theorycited by 218
  2. 2024PrivacyLens: Evaluating Privacy Norm Awareness of Language Models in Actioncited by 161
  3. 2024On Protecting the Data Privacy of Large Language Models (LLMs): A Surveycited by 91
  4. 2024AirGapAgent: Protecting Privacy-Conscious Conversational Agentscited by 90
  5. 2023"It's a Fair Game", or Is It? Examining How Users Navigate Disclosure Risks and Benefits When Using LLM-Based Conversational Agentscited by 60
  6. 2025Privacy in Action: Towards Realistic Privacy Mitigation and Evaluation for LLM-Powered Agentscited by 33
  7. 2024Alternate Preference Optimization for Unlearning Factual Knowledge in Large Language Modelscited by 32
  8. 2023Privacy- and Utility-Preserving NLP with Anonymized Data: A case study of Pseudonymizationcited by 18
  9. 2025SoK: The Privacy Paradox of Large Language Models: Advancements, Privacy Risks, and Mitigationcited by 18
  10. 2022How to keep text private? A systematic review of deep learning methods for privacy-preserving natural language processingcited by 16
  11. 2025Operationalizing Data Minimization for Privacy-Preserving LLM Promptingcited by 8
  12. 2025Beyond Jailbreaking: Auditing Contextual Privacy in LLM Agentscited by 7
3 more
  1. 2017Toward sensitive document release with privacy guaranteescited by 6
  2. 2025User Perceptions vs. Proxy LLM Judges: Privacy and Helpfulness in LLM Responses to Privacy-Sensitive Scenarioscited by 1
  3. 2025Semantically-Aware LLM Agent to Enhance Privacy in Conversational AI Servicescited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Protecting Users From Themselves: Safeguarding Contextual Privacy in Interactions with Conversational Agents
  2. 2026MuPPET: A Benchmark for Contextual Privacy of LLM Assistants in Multi-Party Conversations
  3. 2025MAGPIE: A dataset for Multi-AGent contextual PrIvacy Evaluation
  4. 2026Anonymous-by-Construction: An LLM-Driven Framework for Privacy-Preserving Text
  5. 20251-2-3 Check: Enhancing Contextual Privacy in LLM via Multi-Agent Reasoning
  6. 2025LLM-as-a-Judge for Privacy Evaluation? Exploring the Alignment of Human and LLM Perceptions of Privacy in Textual Data
  7. 2026OCELOT: Inference-Leakage Budgets for Privacy-Preserving LLM Agents
  8. 2025PrivacyBench: A Conversational Benchmark for Evaluating Privacy in Personalized AI