mlsanityData to 7 Oct 2026

On the Evidentiary Limits of Membership Inference for Copyright Auditing

arXiv 2601.12937 ↗First posted 19 Jan 2026cs.CR cs.AIcited by 0

On the map: LLM safety and security → Membership inference attacks

Abstract (arXiv)

As large language models (LLMs) are trained on increasingly opaque corpora, membership inference attacks (MIAs) have been proposed to audit whether copyrighted texts were used during training, despite growing concerns about their reliability under realistic conditions. We ask whether MIAs can serve as admissible evidence in adversarial copyright disputes where an accused model developer may obfuscate training data while preserving semantic content, and formalize this setting through a judge-prosecutor-accused communication protocol. To test robustness under this protocol, we introduce SAGE (Structure-Aware SAE-Guided Extraction), a paraphrasing framework guided by Sparse Autoencoders (SAEs) that rewrites training data to alter lexical structure while preserving semantic content and downstream utility. Our experiments show that state-of-the-art MIAs degrade when models are fine-tuned on SAGE-generated paraphrases, indicating that their signals are not robust to semantics-preserving transformations. While some leakage remains in certain fine-tuning regimes, these results suggest that MIAs are brittle in adversarial settings and insufficient, on their own, as a standalone mechanism for copyright auditing of LLMs.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Membership inference attacks, one of 12 directions in LLM safety and security. Builds on 41 of its 57 references that are in the map. First posted on arXiv 19 Jan 2026.

Builds on

41 of 57 references.

Closest ancestors · 10

Same direction, newest first

  1. 2024SoK: Membership Inference Attacks on LLMs are Rushing Nowhere (and How to Fix It)25 Jun
  2. 2024Blind Baselines Beat Membership Inference Attacks for Foundation Models23 Jun
  3. 2024ReCaLL: Membership Inference via Relative Conditional Log-Likelihoods23 Jun
  4. 2024LLM Dataset Inference: Did you train on my dataset?10 Jun
  5. 2024Do Membership Inference Attacks Work on Large Language Models?12 Feb
  6. 2023Did the Neurons Read your Book? Document-level Membership Inference for Large Language Models23 Oct
  7. 2023Membership Inference Attacks against Language Models via Neighbourhood Comparison29 May
  8. 2022Quantifying Privacy Risks of Masked Language Models Using Membership Inference Attacks8 Mar
  9. 2021Membership Inference Attacks From First Principles7 Dec
  10. 2021On the Importance of Difficulty Calibration in Membership Inference Attacks15 Nov

Other references · 31

  1. 2024The Llama 3 Herd of Modelscited by 14,192
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  3. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  4. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,655
  5. 2016Deep Learning with Differential Privacycited by 2,682
  6. 2020Extracting Training Data from Large Language Modelscited by 2,144
  7. 2022Large Language Models Encode Clinical Knowledgecited by 1,807
  8. 2023Pythia: A Suite for Analyzing Large Language Models Across Training and Scalingcited by 1,734
  9. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  10. 2023Large Language Models are not Fair Evaluatorscited by 1,050
  11. 2023BloombergGPT: A Large Language Model for Financecited by 923
  12. 2023Can Large Language Models Be an Alternative to Human Evaluations?cited by 729
19 more
  1. 2021Large Language Models Can Be Strong Differentially Private Learnerscited by 405
  2. 2023Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defensecited by 401
  3. 2021Differentially Private Fine-tuning of Language Modelscited by 392
  4. 2023Scaling Data-Constrained Language Modelscited by 391
  5. 2023Detecting Pretraining Data from Large Language Modelscited by 375
  6. 2024LoRA Learns Less and Forgets Lesscited by 313
  7. 2022Deduplicating Training Data Mitigates Privacy Risks in Language Modelscited by 309
  8. 2024Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Modelscited by 301
  9. 2023Foundation Models and Fair Usecited by 112
  10. 2024Min-K%++: Improved Baseline for Detecting Pre-Training Data from Large Language Modelscited by 102
  11. 2024LoRA vs Full Fine-tuning: An Illusion of Equivalencecited by 102
  12. 2023The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AIcited by 73
  13. 2024DE-COP: Detecting Copyrighted Content in Language Models Training Datacited by 72
  14. 2019One Epoch Is All You Needcited by 41
  15. 2023Differentially Private Low-Rank Adaptation of Large Language Model Using Federated Learningcited by 41
  16. 2024Copyright Traps for Large Language Modelscited by 36
  17. 2023The Janus Interface: How Fine-Tuning in Large Language Models Amplifies the Privacy Riskscited by 35
  18. 2024Con-ReCall: Detecting Pre-training Data in LLMs via Contrastive Decodingcited by 20
  19. 2024Reconstruct Your Previous Conversations! Comprehensively Investigating Privacy Leakage Risks in Conversations with GPT Modelscited by 16

Built on by

Not cited yet.