In plain words: Game characters powered by language models are given secret backstory details that players should never hear. The study tests whether crafted prompts can trick these characters into revealing those hidden secrets.
Abstract
Large Language Models (LLMs) are increasingly used to generate dynamic dialogue for game NPCs. However, their integration raises new security concerns. In this study, we examine whether adversarial prompt injection can cause LLM-based NPCs to reveal hidden background secrets that are meant to remain undisclosed.
Kyohei Shiomi, Zhuotao Lian, Toru Nakanishi, Teruaki Kitasuka
arXiv:2508.19288 · cs.CR, cs.AI · submitted Aug 25, 2025
abstract · pdf · html
Recently I've tried Baldur's Gate 3, gameplay looks nice, but options in dialogues written by 12 y.o. kid. It would be nice to talk via LLM that would have the described character personality.