Ignore All Previous Instructions: Jailbreaking as a de-escalatory peace building practise to resist LLM social media bots
This addresses the issue of conflict escalation from LLM social media bots for online users, offering a novel countermeasure beyond platform moderation.
The paper tackles the problem of LLM-driven political manipulation on social media by proposing jailbreaking as a user-centric de-escalation practice, where users interact with suspected bot accounts to expose automated behavior and disrupt misleading narratives.
Large Language Models have intensified the scale and strategic manipulation of political discourse on social media, leading to conflict escalation. The existing literature largely focuses on platform-led moderation as a countermeasure. In this paper, we propose a user-centric view of "jailbreaking" as an emergent, non-violent de-escalation practice. Online users engage with suspected LLM-powered accounts to circumvent large language model safeguards, exposing automated behaviour and disrupting the circulation of misleading narratives.