mlsanityData to 7 Oct 2026

Undetectable Watermarks for Language Models

arXiv 2306.09194 ↗First posted 25 May 2023cs.CR cs.CL cs.LGcited by 208

On the map: LLM safety and security → Watermarking generated content and models

Abstract (arXiv)

Recent advances in the capabilities of large language models such as GPT-4 have spurred increasing concern about our ability to detect AI-generated text. Prior works have suggested methods of embedding watermarks in model outputs, by noticeably altering the output distribution. We ask: Is it possible to introduce a watermark without incurring any detectable change to the output distribution? To this end we introduce a cryptographically-inspired notion of undetectable watermarks for language models. That is, watermarks can be detected only with the knowledge of a secret key; without the secret key, it is computationally intractable to distinguish watermarked outputs from those of the original model. In particular, it is impossible for a user to observe any degradation in the quality of the text. Crucially, watermarks should remain undetectable even when the user is allowed to adaptively query the model with arbitrarily chosen prompts. We construct undetectable watermarks based on the existence of one-way functions, a standard assumption in cryptography.

Key facts. Cited by 208 papers in the map (data through 7 Oct 2026). Placed in Watermarking generated content and models, one of 12 directions in LLM safety and security. Builds on 11 of its 44 references that are in the map. First posted on arXiv 25 May 2023.

Builds on

11 of 44 references.

Closest ancestors · 3

Same direction, newest first

  1. 2023Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defense23 Mar
  2. 2023Can AI-Generated Text be Reliably Detected?17 Mar
  3. 2023A Watermark for Large Language Models24 Jan

Other references · 8

  1. 2023DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvaturecited by 705
  2. 2019Defending Against Neural Fake Newscited by 553
  3. 2019GLTR: Statistical Detection and Visualization of Generated Textcited by 490
  4. 2023GPT detectors are biased against non-native English writerscited by 175
  5. 2020Automatic Detection of Machine Generated Text: A Critical Surveycited by 160
  6. 2020Adversarial Watermarking Transformer: Towards Tracing Text Provenance with Data Hidingcited by 121
  7. 2023On the Possibilities of AI-Generated Text Detectioncited by 88
  8. 2022Planting Undetectable Backdoors in Machine Learning Modelscited by 55

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 24; 2024: 60; 2025: 72; 2026: 52.

208 papers.

Sort:

By area:

  1. 2023Robust Distortion-free Watermarks for Language Modelscited by 279
  2. 2023Provable Robust Watermarking for AI-Generated Textcited by 259
  3. 2023A Semantic Invariant Robust Watermark for Large Language Modelscited by 97
  4. 2023Identifying and Mitigating the Security Risks of Generative AIcited by 90
  5. 2025Subliminal Learning: Language models transmit behavioral traits via hidden signals in datacited by 89
  6. 2023Three Bricks to Consolidate Watermarks for Large Language Modelscited by 80
  7. 2023Watermarks in the Sand: Impossibility of Strong Watermarking for Generative Modelscited by 77
  8. 2023Use of LLMs for Illicit Purposes: Threats, Prevention Measures, and Vulnerabilitiescited by 76
  9. 2024An Undetectable Watermark for Generative Image Modelscited by 74
  10. 2024Detecting Multimedia Generated by Large AI Models: A Surveycited by 73
  11. 2023On the Learnability of Watermarks for Language Modelscited by 72
  12. 2023Publicly-Detectable Watermarking for Language Modelscited by 69
38 more
  1. 2023REMARK-LLM: A Robust and Efficient Watermarking Framework for Generative Large Language Modelscited by 64
  2. 2023Mark My Words: Analyzing and Evaluating Language Model Watermarkscited by 62
  3. 2024Adaptive Text Watermark for Large Language Modelscited by 60
  4. 2023An Unforgeable Publicly Verifiable Watermark for Large Language Modelscited by 54
  5. 2024Watermarking Makes Language Models Radioactivecited by 52
  6. 2023Advancing Beyond Identification: Multi-bit Watermark for Large Language Modelscited by 51
  7. 2024WaterMax: breaking the LLM watermark detectability-robustness-quality trade-offcited by 48
  8. 2023A Resilient and Accessible Distribution-Preserving Watermark for Large Language Modelscited by 47
  9. 2024No Free Lunch in LLM Watermarking: Trade-offs in Watermarking Design Choicescited by 47
  10. 2024SoK: Watermarking for AI-Generated Contentcited by 43
  11. 2024Pseudorandom Error-Correcting Codescited by 40
  12. 2023Fighting Fire with Fire: Can ChatGPT Detect AI-generated Text?cited by 39
  13. 2024Copyright Protection in Generative AI: A Technical Perspectivecited by 35
  14. 2024Hey, That's My Model! Introducing Chain & Hash, An LLM Fingerprinting Techniquecited by 34
  15. 2023HuRef: HUman-REadable Fingerprint for Large Language Modelscited by 32
  16. 2023Towards Possibilities & Impossibilities of AI-generated Text Detection: A Surveycited by 28
  17. 2025SynthID-Image: Image watermarking at internet scalecited by 26
  18. 2023From Text to Source: Results in Detecting Large Language Model-Generated Contentcited by 25
  19. 2023Towards Optimal Statistical Watermarkingcited by 25
  20. 2023Embarrassingly Simple Text Watermarkscited by 23
  21. 2024Deconstructing The Ethics of Large Language Models from Long-standing Issues to New-emerging Dilemmas: A Surveycited by 23
  22. 2024Detecting AI-Generated Text: Factors Influencing Detectability with Current Methodscited by 23
  23. 2024Black-Box Detection of Language Model Watermarkscited by 22
  24. 2024Bileve: Securing Text Provenance in Large Language Models Against Spoofing with Bi-level Signaturecited by 21
  25. 2025AdaDetectGPT: Adaptive Detection of LLM-Generated Text with Statistical Guaranteescited by 20
  26. 2024Learning to Watermark LLM-generated Text via Reinforcement Learningcited by 19
  27. 2024Double-I Watermark: Protecting Model Copyright for LLM Fine-tuningcited by 18
  28. 2024Debiasing Watermarks for Large Language Models via Maximal Couplingcited by 18
  29. 2024Permute-and-Flip: An optimally stable and watermarkable decoder for LLMscited by 17
  30. 2024AI Risk Management Should Incorporate Both Safety and Securitycited by 17
  31. 2024Theoretically Grounded Framework for LLM Watermarking: A Distribution-Adaptive Approachcited by 17
  32. 2025CoheMark: A Novel Sentence-Level Watermark for Enhanced Text Qualitycited by 17
  33. 2023Necessary and Sufficient Watermark for Large Language Modelscited by 16
  34. 2024Watermarking Language Models for Many Adaptive Userscited by 16
  35. 2025Improving Your Model Ranking on Chatbot Arena by Vote Riggingcited by 16
  36. 2023On the Trustworthiness Landscape of State-of-the-art Generative Models: A Survey and Outlookcited by 15
  37. 2024Can Watermarking Large Language Models Prevent Copyrighted Text Generation and Hide Training Data?cited by 15
  38. 2024Ward: Provable RAG Dataset Inference via LLM Watermarkscited by 15

50 most cited of 208 All 208 (JSON)