Rescriber: Smaller-LLM-Powered User-Led Data Minimization for LLM-Based Chatbots
Open accessThe DOI matches its Crossref record; 43 of the 43 references it cites were found in a registry or an index.
Abstract
The proliferation of LLM-based conversational agents has resulted in excessive disclosure of identifiable or sensitive information.However, existing technologies fail to offer perceptible control or account for users' personal preferences about privacy-utility tradeoffs due to the lack of user involvement.To bridge this gap, we designed, built, and evaluated Rescriber, a browser extension that supports user-led data minimization in LLM-based conversational agents by helping users detect and sanitize personal information in their prompts.Our studies (N=Rescriber) showed that Rescriber helped users reduce unnecessary disclosure and addressed their privacy concerns.Users' subjective perceptions of the system powered by Llama3-8B were on par with that by GPT-4o.The comprehensiveness and consistency of the detection and sanitization emerge as essential factors that affect users' trust and perceived protection.Our findings confirm the viability of smaller-LLM-powered, userfacing, on-device privacy controls, presenting a promising approach to address the privacy and trust challenges of AI.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Figures
![Figure 1 A snapshot of Rescriber user experience when using ChatGPT for data analysis. Rescriber displays a tooltip that highlights the detected personal information in the user's message (A), and offers a control panel (B) where users can either replace the information with a placeholder (B1), abstract the information to a more general version (B2), or revert the actions (B3). All the changes would be immediately effective on the message that would be sent out in the input box (C). In this example, the user replaced the detected names (e.g., James Williams) to corresponding placeholders (e.g., [NAME4]). For the sent messages (D1) and ChatGPT's responses (D2), Rescriber replaces the placeholders with the original PIIs to increase readability and facilitate copying. When the user mouses over the highlighted PIIs, the placeholders used in the actual messages sent and received are revealed (D).](https://figures.cochat.ai/paper-figures/131ba56bf7ef44bcbb89fc8228619a78/rescriber-smaller-llm-powered-user-led-data-minimization-for-llm-based-chatbots-figure-1.png)




Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
LLaMA: Open and Efficient Foundation Language Models
Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.-A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F., et al. (2023) · arXiv · doi:10.48550/arxiv.2302.13971 · Cited by 4,602
Internet Users' Information Privacy Concerns (IUIPC): The Construct, the Scale, and a Causal Model
Malhotra, N. K., Kim, S. S., & Agarwal, J. (2004) · Information Systems Research · doi:10.1287/isre.1040.0032 · Cited by 3,173 · cited 1×
“Additionally, providing granular control also helps users better minimize unnecessary disclosure by striking a balance between privacy and utility.”
(2011) · IEEE Software · doi:10.1109/ms.2011.67 · Cited by 1,989
Sparks of Artificial General Intelligence: Early experiments with GPT-4
Bubeck, S., Chandrasekaran, V., Eldan, R., Gehrke, J., Horvitz, E., Kamar, E., Lee, P., Lee, Y. T., Li, Y., Lundberg, S., et al. (2023) · arXiv · doi:10.48550/arxiv.2303.12712 · Cited by 1,526
Emergent Abilities of Large Language Models
Wei, J., Tay, Y., Bommasani, R., Raffel, C., Zoph, B., Borgeaud, S., Yogatama, D., Bosma, M., Zhou, D., Metzler, D., et al. (2022) · arXiv · doi:10.48550/arxiv.2206.07682 · Cited by 1,264
Schaar, P. (2010) · Identity in the Information Society · doi:10.1007/s12394-010-0055-x · Cited by 387
Beyer, H., & Holtzblatt, K. (1999) · Interactions · doi:10.1145/291224.291229 · Cited by 367
A taxonomy of privacy-preserving record linkage techniques
Vatsalan, D., Christen, P., & Verykios, V. S. (2013) · Information Systems · doi:10.1016/j.is.2012.11.005 · Cited by 295 · cited 1×
“” Concerns about identifiability also extended beyond direct identifiers to demographic details like gender, raising fears of linkage attacks [50] and broader profiles emerging from multiple data points.”
UMUX-LITE: when there's no time for the SUS
Lewis, J. R., Utesch, B. S., & Maher, D. E. (2013) · Proceedings of the SIGCHI Conference on Human Factors in Computing Systems · doi:10.1145/2470654.2481287 · Cited by 270
PrivacyLens: Evaluating Privacy Norm Awareness of Language Models in ActionPreprint
Shao, Y., Li, T.-S., Shi, W.-Y., Liu, Y., & Yang, D. (2024) · Neural Information Processing Systems · doi:10.48550/arxiv.2409.00138 · Cited by 261
Self-citation
Research works printed but not linked: 10
- 12. Youngwook Do, Nivedita Arora, Ali Mirzazadeh, Injoo Moon, Eryue Xu, Zhi- han Zhang, Gregory D Abowd, and Sauvik Das. 2023. Powering for privacy: improving user trust in smart speaker microphones with intentional powering and perceptible assurance. In 32nd USENIX Security Symposium (USENIX Security 23). 2473-2490.
- 16. James W Gentry. 1990. What is experiential learning. Guide to business gaming and experiential learning 9, 1 (1990), 20-32.
- 19. Nikhil Kandpal, Eric Wallace, and Colin Raffel. 2022. Deduplicating training data mitigates privacy risks in language models. In International Conference on Machine Learning. PMLR, 10697-10707.
- 31. Niloofar Mireshghallah, Maria Antoniak, Yash More, Yejin Choi, and Golnoosh Farnadi. 2024. Trust No Bot: Discovering Personal Disclosures in Human-LLM Conversations in the Wild. arXiv preprint arXiv:2407.11438 (2024). · arXiv
- 34. Helen Nissenbaum. 2004. Privacy as contextual integrity. Wash. L. Rev. 79 (2004), 119.
- 35. Annika Willoch Olstad, Anthi Papadopoulou, and Pierre Lison. 2023. Generation of Replacement Options in Text Sanitization. In Proceedings of the 24th Nordic Conference on Computational Linguistics (NoDaLiDa). 292-300.
- 39. Bashir Rastegarpanah, Krishna Gummadi, and Mark Crovella. 2021. Auditing black-box prediction models for data minimization compliance. Advances in Neural Information Processing Systems 34 (2021), 20621-20632.
- 42. Rylan Schaeffer, Brando Miranda, and Sanmi Koyejo. 2024. Are emergent abilities of large language models a mirage? Advances in Neural Information Processing Systems 36 (2024).
- 45. Tanusree Sharma, Lin Kyi, Yang Wang, and Asia J Biega. 2024. " I'm not con- vinced that they don't collect more than is necessary":{User-Controlled} Data Minimization Design in Search Engines. In 33rd USENIX Security Symposium (USENIX Security 24). 2797-2812.
- 58. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2024. Judging LLM-as-a-judge with MT- bench and Chatbot Arena. In Proceedings of the 37th International Conference on Neural Information Processing Systems (New Orleans, LA, USA) (NIPS '23).
Cited by
Luo, H., Liu, Y., Zhang, R., Wang, J., Sun, G., Niyato, D. T., Yu, H., Xiong, Z., Wang, X., & Shen, X. (2025) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2025.3612760 · Cited by 94
Chen, C., Zhou, D., Ye, Y., Li, T. J.-J., & Yao, Y. (2025) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/3708359.3712156 · Cited by 64
Self-citationUnderstanding Users' Privacy Perceptions Towards LLM's RAG-based Memory
Zhang, S., Ma, R., Ma, Y., Li, S., Xu, Y., Yi, X., & Li, H. (2025) · Proceedings of the 2025 Workshop on Human-Centered AI Privacy and Security · doi:10.1145/3733816.3760750 · Cited by 6
“Some operate at the data interface: OneShield [2] filters both user inputs and model outputs, while Rescriber [50] leverages an LLM to minimize sensitive content in queries in real time.”
Sharma, T., Tseng, Y., Zhang, L., Ide, A., Mack, K., Findlater, L., Gurari, D., & Wang, Y. (2025) · International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/3663547.3746335 · Cited by 19
Towards Aligning Personalized AI Agents with Users' Privacy Preference
Zhang, S., Ma, Y., Chen, J., Li, S., Yi, X., & Li, H. (2025) · doi:10.1145/3733816.3760752 · Cited by 20
Zhang, Z., Bao, C., Pan, X., Chang, C.-M., Igarashi, T., & Zhang, G. (2025) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3706599.3720234 · Cited by 18
Gao, H.-A., Geng, J., Hua, W., Hu, M., Juan, X., Liu, H., Liu, S., Qiu, J., Qi, X., Wu, Y., et al. (2025) · arXiv · doi:10.48550/arxiv.2507.21046 · Cited by 3
Self-citationSensorimotor Regularities as Alignment between Humans and Large Language Models
Li, J., Hu, J., & Kristensson, P. O. (2026) · ACM Transactions on Computer-Human Interaction · doi:10.1145/3778353 · Cited by 1
Psarra, E., & Stefanidis, K. G. (2026) · Applied Sciences · doi:10.3390/app16073332 · Cited by 2
Towards Usable, Privacy Respecting Long-Term Memory for LLM-based Conversational Agents
Malki, L. M. (2026) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3772363.3799198 · Cited by 3
Wang, K., Shen, Z., Zhang, Y., Cheung, M. M., Luo, X., Ngai, G., & Fu, E. Y. (2025) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3746027.3758164 · Cited by 3
Zhang, S., Ma, Y., Hu, Y. '., Dang, T., Jia, H., Yi, X., & Li, H. (2025) · Companion of the 2025 ACM International Joint Conference on Pervasive and Ubiquitous Computing · doi:10.1145/3714394.3756335
Privi: Assisting Users in Authoring Contextual Privacy Rules with an LLM Sandbox
Guo, B., Zhang, Z., & Li, T. (2025) · doi:10.1145/3733816.3760756 · Cited by 2
Self-citationWang, H., Winkle, K., & Castellano, G. (2026) · ACM/IEEE International Conference on Human-Robot Interaction (HRI) · doi:10.1145/3776734.3794552 · Cited by 2
A Unified Framework for Quantifying and Reducing Persuasion within a TextPreprint
Mura, D. A., Loddo, A., Sanguinetti, M., Atzori, M., & Ruberto, C. D. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6502554
User Autonomy in Human-LLM Interaction: A Scoping Review
Wang, Y., & Wang, G. (2026) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3772363.3798855 · Cited by 2
The Privacy Paradox of LLMs: User Perceptions and the Reality of PII Leakage
Cheng, S., Xu, H., Meng, S., Hao, S., Yue, C., & Li, Z. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3791809 · Cited by 2
Wang, G., & Huang, F. (2026) · Journal of Strategic Marketing · doi:10.1080/0965254x.2026.2675435 · Cited by 1
Can Large Language Models Really Recognize Your Name?
Pham, D., Kairouz, P., Mireshghallah, F., Bagdasarian, E., Pham, C. M., & Houmansadr, A. (2026) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3805689.3806747 · Cited by 8
Do Attachment Styles Shape ChatGPT Usage?
Wang, M., Li, J. X., Ngo, S., Davis, K. S., & Hiniker, A. (2026) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3805689.3812418 · Cited by 1
WAIT-A-SEC: A Human-Centered Design Approach to Privacy Communication for Social Media Oversharing
Kelly, T., & Rao, P. R. (2026) · ACM International Conference on Design of Communication (SIGDOC) · doi:10.1145/3772063.3841664
Mediação de Privacidade para Interações com Modelos de Linguagem de Grande Porte
da Silva, F. D. L., & Coleti, T. A. (2026) · Anais do Workshop sobre as Implicações da Computação na Sociedade (WICS) · doi:10.5753/wics.2026.23591
Gupta, S., Abdollahi, S., Rahman, M. M., Shin, M. M., & Khan, Y. (2026) · De Gruyter eBooks · doi:10.1515/9783111478012-011
LLM Firewall: Private LLM Program Repair Using Hybrid Intent Inference
Akyol, G., Latifoğlu, İ., Kaya, A., & Tüzün, E. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-36590-3_25
Kitahata, T., Seki, K., & Nadamoto, A. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-5696-9_28
Metadata
- Authors
- Zhou, J., Xu, E., Wu, Y., & Li, T.
- Affiliations
- Northeastern University (US)
- Venue
- Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems
- Publisher
- ACM
- Published
- 2025-04-25
- Pages
- 1-28
- OpenAlex
- W4409870977
- Semantic Scholar
- f1366e505de4f1d0e901903e3c17471033758a96
- Cited by
- 17
- References
- 23
- Citations by year
- 2026: 12 · 2025: 5
- FWCI
- 57.11
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- CHI Conference on Human Factors in Computing Systems (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- Topics
- Social Sciences › Sociology and Political Science › Privacy, Security, and Data Protection
- Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Keywords
- LLM-based chatbots, data minimization, user privacy control, personal information detection, prompt sanitization, browser extension, small language models, privacy-utility trade-off, user trust, sensitive information disclosure, Llama-3-8B, GPT-4o
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.