Automating Deception: Scalable Multi-Turn LLM Jailbreaks

TL;DR

This paper presents an automated method to generate large-scale multi-turn jailbreak datasets exploiting psychological principles, revealing significant vulnerabilities in most LLMs and highlighting the need for improved safety defenses.

Contribution

It introduces a scalable, automated pipeline for creating psychologically-grounded multi-turn jailbreak datasets and systematically evaluates model robustness against these attacks.

Findings

GPT models are highly vulnerable to multi-turn attacks with up to 32% increase in success rate.

Google's Gemini 2.5 Flash shows near immunity to multi-turn jailbreaks.

Model robustness varies significantly across different LLM families.

Abstract

Multi-turn conversational attacks, which leverage psychological principles like Foot-in-the-Door (FITD), where a small initial request paves the way for a more significant one, to bypass safety alignments, pose a persistent threat to Large Language Models (LLMs). Progress in defending against these attacks is hindered by a reliance on manual, hard-to-scale dataset creation. This paper introduces a novel, automated pipeline for generating large-scale, psychologically-grounded multi-turn jailbreak datasets. We systematically operationalize FITD techniques into reproducible templates, creating a benchmark of 1,500 scenarios across illegal activities and offensive content. We evaluate seven models from three major LLM families under both multi-turn (with history) and single-turn (without history) conditions. Our results reveal stark differences in contextual robustness: models in the GPT family demonstrate a significant vulnerability to conversational history, with Attack Success Rates (ASR) increasing by as much as 32 percentage points. In contrast, Google's Gemini 2.5 Flash exhibits exceptional resilience, proving nearly immune to these attacks, while Anthropic's Claude 3 Haiku shows strong but imperfect resistance. These findings highlight a critical divergence in how current safety architectures handle conversational context and underscore the need for defenses that can resist narrative-based manipulation.