Publication Type

Conference Proceeding Article

Version

publishedVersion

Publication Date

7-2026

Abstract

Large language models (LLMs) are increasingly trained on massive, heterogeneous text corpora, raising serious concerns about the unauthorised use of proprietary or personal data during model training. In this work, we address the problem of data protection against unwanted model learning in a realistic blackbox setting. We propose Disclaimer Injection, a novel data-level defence that renders text unlearnable to LLMs. Rather than relying on model-side controls or explicit data removal, our approach exploits the models’ own alignment mechanisms: injecting carefully designed alignment-triggers to prevent effective learning. Through layer-wise analysis, we find that finetuning on such protected data induces persistent activation of alignment-related layers, causing alignment constraints to override task learning even on common inputs. Consequently, models trained on such data exhibit substantial and systematic performance degradation compared to standard fine-tuning. Our results identify alignment behaviour as a previously unexplored lever for data protection and, to our knowledge, present the first practical method for restricting data learnability at LLM scale without requiring access to or modification of the training pipeline. Disclaimer: This paper contains potentially harmful content.

Discipline

Artificial Intelligence and Robotics | Information Security

Research Areas

Software and Cyber-Physical Systems

Areas of Excellence

Digital transformation

Publication

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), San Diego, California, July 2-7

First Page

40587

Last Page

40598

Publisher

ACL

City or Country

San Diego, California

Download

Share

COinS