Privacy Ripple Effects from Adding or Removing Personal Information in Language Model Training

TL;DR

This paper investigates how adding or removing personal information from training data affects the memorization of PII in language models, revealing dynamic privacy risks during training.

Contribution

It identifies and characterizes three novel phenomena showing how PII memorization evolves and interacts during model training and dataset updates.

Findings

Assisted memorization can cause earlier PII to be recalled later in training.

Adding PII can significantly increase memorization of other PII.

Removing PII can unexpectedly lead to other PII being memorized.

Abstract

Due to the sensitive nature of personally identifiable information (PII), its owners may have the authority to control its inclusion or request its removal from large-language model (LLM) training. Beyond this, PII may be added or removed from training datasets due to evolving dataset curation techniques, because they were newly scraped for retraining, or because they were included in a new downstream fine-tuning stage. We find that the amount and ease of PII memorization is a dynamic property of a model that evolves throughout training pipelines and depends on commonly altered design choices. We characterize three such novel phenomena: (1) similar-appearing PII seen later in training can elicit memorization of earlier-seen sequences in what we call assisted memorization, and this is a significant factor (in our settings, up to 1/3); (2) adding PII can increase memorization of other PII significantly (in our settings, as much as $\approx\!7.5\times$); and (3) removing PII can lead to other PII being memorized. Model creators should consider these first- and second-order privacy risks when training models to avoid the risk of new PII regurgitation.