mlsanityData to 14 Oct 2026

Privacy Policy Enforcement Guardrails for Data-Sensitive Retrieval-Augmented Generation

arXiv 2605.17034 ↗First posted 16 May 2026cs.LG cs.AI cs.CRcited by 0

On the map: Retrieval-augmented generation → Privacy and security of RAG

also in 1 more
Abstract (arXiv)

Standard PII filters often miss contextual data leakage in RAG systems, such as non-regulated attribute clusters that collectively identify individuals. We introduce a Privacy Policy Enforcement (PPE) framework using dual one-class density estimators with fused text embeddings and a calibrated abstain region for out-of-distribution inputs. Using an axis-stratified, multi-LLM synthetic data pipeline across medicine, finance, and law, we found that traditional Gaussian Mixture baselines fail on borderline-safe stress tests by focusing on linguistic register rather than content. Our proposed T3+OCSVM detector, trained on safe and borderline-safe data, achieves a borderline AUROC of 0.93+ while reducing false positives by 44-55 percentage points and maintaining millisecond latency. Compared to supervised MLP classifiers or 14B-parameter LLM judges, our framework offers superior operational suitability, as the former suffers from high abstention rates and the latter from latency and calibration issues. This methodology provides a robust stress-testing standard for any synthetic-data-trained classifier.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Privacy and security of RAG, one of 15 directions in Retrieval-augmented generation. Builds on 32 of its 52 references that are in the map. First posted on arXiv 16 May 2026.

Builds on

32 of 52 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Generating Is Believing: Membership Inference Attacks against Retrieval-Augmented Generation27 Jun

Other references · 31

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  3. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,695
  4. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  5. 2023Toolformer: Language Models Can Teach Themselves to Use Toolscited by 4,182
  6. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 3,104
  7. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 3,050
  8. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,281
  9. 2020Extracting Training Data from Large Language Modelscited by 2,189
  10. 2020REALM: Retrieval-Augmented Language Model Pre-Trainingcited by 2,141
  11. 2023Jailbroken: How Does LLM Safety Training Fail?cited by 1,651
  12. 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,207
19 more
  1. 2022Quantifying Memorization Across Neural Language Modelscited by 763
  2. 2021FinQA: A Dataset of Numerical Reasoning over Financial Datacited by 562
  3. 2024WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMscited by 496
  4. 2019Likelihood Ratios for Out-of-Distribution Detectioncited by 453
  5. 2023Scalable Extraction of Training Data from (Production) Language Modelscited by 434
  6. 2023NeMo Guardrails: A Toolkit for Controllable and Safe LLM Applications with Programmable Railscited by 365
  7. 2023Analyzing Leakage of Personally Identifiable Information in Language Modelscited by 285
  8. 2023FinanceBench: A New Benchmark for Financial Question Answeringcited by 270
  9. 2023Membership Inference Attacks against Language Models via Neighbourhood Comparisoncited by 258
  10. 2019A Question-Entailment Approach to Question Answeringcited by 134
  11. 2020Density of States Estimation for Out-of-Distribution Detectioncited by 74
  12. 2022EUR-Lex-Sum: A Multi- and Cross-lingual Dataset for Long-form Summarization in the Legal Domaincited by 40
  13. 2024Forte : Finding Outliers with Representation Typicality Estimationcited by 9
  14. 2025K4: Online Log Anomaly Detection Via Unsupervised Typicality Learningcited by 6
  15. 2025LABELING COPILOT: A Deep Research Agent for Automated Data Curation in Computer Visioncited by 6
  16. 2026Trust The Typicalcited by 3
  17. 2025ZKPROV: A Zero-Knowledge Approach to Dataset Provenance for Large Language Modelscited by 2
  18. 2025Exploring Membership Inference Vulnerabilities in Clinical Large Language Modelscited by 1
  19. 2026Context Determines Optimal Architecture in Materials Segmentationcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024SafeSynthDP: Leveraging Large Language Models for Privacy-Preserving Synthetic Data Generation Using Differential Privacy
  2. 2026LLM-CEG: Extending the Classification Error Gauge Framework for Privacy Auditing of Large Language Models ↗
  3. 2026Advancing the State-of-the-Art in Empirical Privacy Auditing
  4. 2025PrivaCI-Bench: Evaluating Privacy with Contextual Integrity and Legal Compliance
  5. 2025Transparent NLP: Using RAG and LLM Alignment for Privacy Q&A
  6. 2026LADFA: A Framework of Using Large Language Models and Retrieval-Augmented Generation for Personal Data Flow Analysis in Privacy Policies
  7. 2025Privacy-Aware Decoding: Mitigating Privacy Leakage of Large Language Models in Retrieval-Augmented Generation
  8. 2024Privacy Policy Analysis through Prompt Engineering for LLMs