mlsanityData to 14 Oct 2026

Machine Unlearning Doesn't Do What You Think: Lessons for Generative AI Policy and Research

arXiv 2412.06966 ↗First posted 9 Dec 2024cs.LG cs.AI cs.CYcited by 15

On the map: LLM safety and security → Machine unlearning

Abstract (arXiv)

"Machine unlearning" is a popular proposed solution for mitigating the existence of content in an AI model that is problematic for legal or moral reasons, including privacy, copyright, safety, and more. For example, unlearning is often invoked as a solution for removing the effects of specific information from a generative-AI model's parameters, e.g., a particular individual's personal data or the inclusion of copyrighted content in the model's training data. Unlearning is also proposed as a way to prevent a model from generating targeted types of information in its outputs, e.g., generations that closely resemble a particular individual's data or reflect the concept of "Spiderman." Both of these goals--the targeted removal of information from a model and the targeted suppression of information from a model's outputs--present various technical and substantive challenges. We provide a framework for ML researchers and policymakers to think rigorously about these challenges, identifying several mismatches between the goals of unlearning and feasible implementations. These mismatches explain why unlearning is not a general-purpose solution for circumscribing generative-AI model behavior in service of broader positive impact.

Key facts. Cited by 15 papers in the map (data through 14 Oct 2026). Placed in Machine unlearning, one of 12 directions in LLM safety and security. Builds on 76 of its 159 references that are in the map. First posted on arXiv 9 Dec 2024.

Builds on

76 of 159 references.

Closest ancestors · 22

Same direction, newest first

  1. 2024Unlearn and Burn: Adversarial Machine Unlearning Requests Destroy Model Accuracy12 Oct
  2. 2024An Adversarial Perspective on Machine Unlearning for AI Safety26 Sep
  3. 2024UnUnlearning: Unlearning is not sufficient for content regulation in advanced generative AI27 Jun
  4. 2024Towards Scalable Exact Machine Unlearning Using Parameter-Efficient Fine-Tuning24 Jun
  5. 2024SOUL: Unlocking the Power of Second-Order Optimization for LLM Unlearning28 Apr
  6. 2024Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning8 Apr
  7. 2024Guardrail Baselines for Unlearning in LLMs5 Mar
  8. 2024Eight Methods to Evaluate Robust Unlearning in LLMs26 Feb
  9. 2024Towards Safer Large Language Models through Machine Unlearning15 Feb
  10. 2024TOFU: A Task of Fictitious Unlearning for LLMs11 Jan
  11. 2023Unlearn What You Want to Forget: Efficient Unlearning for LLMs31 Oct
  12. 2023Large Language Model Unlearning14 Oct
10 more
  1. 2023Who's Harry Potter? Approximate Unlearning in LLMs3 Oct
  2. 2023What can we learn from Data Leakage and Unlearning for Law?19 Jul
  3. 2023Right to be Forgotten in the Era of Large Language Models: Implications, Challenges, and Solutions8 Jul
  4. 2023Machine Unlearning: its nature, scope, and importance for a "delete culture"24 May
  5. 2022Privacy Adhering Machine Un-learning in NLP19 Dec
  6. 2022Control, Confidentiality, and the Right to be Forgotten14 Oct
  7. 2022Knowledge Unlearning for Mitigating Privacy Risks in Language Models4 Oct
  8. 2022Quark: Controllable Text Generation with Reinforced Unlearning26 May
  9. 2022Zero-Shot Machine Unlearning14 Jan
  10. 2021Certifiable Machine Unlearning for Linear Models29 Jun

Other references · 54

  1. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  2. 2022Emergent Abilities of Large Language Modelscited by 2,512
  3. 2022Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language modelscited by 1,998
  4. 2019Language Models as Knowledge Bases?cited by 1,936
  5. 2016Concrete Problems in AI Safetycited by 1,642
  6. 2020Transformer Feed-Forward Layers Are Key-Value Memoriescited by 1,336
  7. 2019Machine Unlearningcited by 1,061
  8. 2023Progress measures for grokking via mechanistic interpretabilitycited by 839
  9. 2022Quantifying Memorization Across Neural Language Modelscited by 763
  10. 2023Towards Automated Circuit Discovery for Mechanistic Interpretabilitycited by 724
  11. 2013An Empirical Investigation of Catastrophic Forgetting in Gradient-Based Neural Networkscited by 722
  12. 2021Knowledge Neurons in Pretrained Transformerscited by 601
38 more
  1. 2019Certified Data Removal from Machine Learning Modelscited by 535
  2. 2023Are Emergent Abilities of Large Language Models a Mirage?cited by 523
  3. 2019Making AI Forget You: Data Deletion in Machine Learningcited by 465
  4. 2019Does Learning Require Memorization? A Short Tale about a Long Tailcited by 460
  5. 2022Memory-Based Model Editing at Scalecited by 447
  6. 2023Scalable Extraction of Training Data from (Production) Language Modelscited by 434
  7. 2020Fine-Tuning Pretrained Language Models: Weight Initializations, Data Orders, and Early Stoppingcited by 384
  8. 2020Descent-to-Delete: Gradient-Based Methods for Machine Unlearningcited by 311
  9. 2023Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theorycited by 230
  10. 2022What Does it Mean for a Language Model to Preserve Privacy?cited by 195
  11. 2018The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigationcited by 165
  12. 2021Understanding the Capabilities, Limitations, and Societal Impact of Large Language Modelscited by 160
  13. 2023Sociotechnical Safety Evaluation of Generative AI Systemscited by 157
  14. 2024Tamper-Resistant Safeguards for Open-Weight LLMscited by 138
  15. 2022The Right to be Forgotten in Federated Learning: An Efficient Realization with Rapid Retrainingcited by 118
  16. 2018Identifying and Controlling Important Neurons in Neural Machine Translationcited by 111
  17. 2024Red-Teaming for Generative AI: Silver Bullet or Security Theater?cited by 83
  18. 2024On the Societal Impact of Open Foundation Modelscited by 65
  19. 2020Learn to Forget: Machine Unlearning via Neuron Maskingcited by 62
  20. 2020Towards Probabilistic Verification of Machine Unlearningcited by 50
  21. 2022Accountability in an Algorithmic Society: Relationality, Responsibility, and Robustness in Machine Learningcited by 48
  22. 2024Machine Unlearning for Image-to-Image Generative Modelscited by 47
  23. 2023Talkin' 'Bout AI Generation: Copyright and the Generative-AI Supply Chaincited by 46
  24. 2024Evaluating Copyright Takedown Methods for Language Modelscited by 43
  25. 2024AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policiescited by 37
  26. 2023Can large language models democratize access to dual-use biotechnology?cited by 35
  27. 2023Arbitrariness and Social Prediction: The Confounding Role of Variance in Fair Classificationcited by 32
  28. 2024Adversaries Can Misuse Combinations of Safe Modelscited by 32
  29. 2024Evaluating Generative AI Systems is a Social Science Measurement Challengecited by 23
  30. 2024The Files are in the Computer: On Copyright, Memorization, and Generative AIcited by 21
  31. 2024International Scientific Report on the Safety of Advanced AI (Interim Report)cited by 19
  32. 2024Breach By A Thousand Leaks: Unsafe Information Leakage in `Safe' AI Responsescited by 18
  33. 2015A Theory of Formal Synthesis via Inductive Learningcited by 17
  34. 2023CommonCanvas: An Open Diffusion Model Trained with Creative-Commons Imagescited by 16
  35. 2023Report of the 1st Workshop on Generative AI and Lawcited by 13
  36. 2024Acceptable Use Policies for Foundation Modelscited by 12
  37. 2024An AI System Evaluation Framework for Advancing AI Safety: Terminology, Taxonomy, Lifecycle Mappingcited by 8
  38. 2022Non-Determinism and the Lawlessness of Machine Learning Codecited by 7

50 most cited of 54 All 54 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 5; 2026: 10.

15 papers.

Sort:

By area:

  1. 2025AegisLLM: Scaling Agentic Systems for Self-Reflective Defense in LLM Securitycited by 10
  2. 2025Unlearning's Blind Spots: Over-Unlearning and Prototypical Relearning Attackcited by 4
  3. 2025Forget to Know, Remember to Use: Context-Aware Unlearning for Large Language Modelscited by 3
  4. 2026Position: The Term "Machine Unlearning" Is Overused in LLMscited by 3
  5. 2026Multilingual Unlearning in LLMs: Transfer, Dynamics, and Reversibilitycited by 2
  6. 2025Unlearning at Scale: State-Exact Trace-Preserving Deletion in Billion-Parameter Language Modelscited by 1
  7. 2026Beyond Forgetting: Representation Misdirection Elicits Controllable Side Behaviors and Capabilitiescited by 1
  8. 2026Temper-Then-Tilt: Principled Unlearning for Generative Models through Tempering and Classifier Guidancecited by 1
  9. 2026Estimating near-verbatim extraction risk in language models with decoding-constrained beam searchcited by 1
  10. 2025Conformal Unlearning: A New Paradigm for Unlearning in Conformal Predictorscited by 0
  11. 2026Form and Function: Machine Unlearning as a Problem of Misaligned Statescited by 0
  12. 2026InstantForget: Update-Free Backdoor Unlearning with Inference-Time Feature Resetcited by 0
3 more
  1. 2026Position: Reasoning is a Learnable Rule-Based Processcited by 0
  2. 2026An Investigation of the NeurIPS and ICML 2025 Position Trackscited by 0
  3. 2026Unmerge: Efficient Machine Unlearning via Task Arithmeticcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024On the Limitations and Prospects of Machine Unlearning for Generative AI
  2. 2024Rethinking Machine Unlearning for Large Language Models
  3. 2025Generative Model Unlearning: A Survey through Target Events, Unlearning Operators, and Evaluation Protocols
  4. 2024Machine Unlearning in Generative AI: A Survey
  5. 2023Exploring the Landscape of Machine Unlearning: A Comprehensive Survey and Taxonomy
  6. 2024The Frontier of Data Erasure: Machine Unlearning for Large Language Models
  7. 2025Model Collapse Is Not a Bug but a Feature in Machine Unlearning for LLMs
  8. 2023Learn to Unlearn: A Survey on Machine Unlearning