This research introduces M2S (Multi-turn-to-Single-turn), a method that condenses complex multi-turn jailbreak attacks into single-prompt vulnerabilities, dramatically increasing attack efficiency and scalability.

• Systematically converts labor-intensive multi-turn attacks into one-shot prompts
• Demonstrates how even sophisticated LLM safety guardrails remain vulnerable
• Shows the unexpected transferability of these attacks across different models
• Provides important insights for improving LLM defense mechanisms

Security Implications: This work reveals concerning security gaps in current LLM safeguards by showing how attacks can be automated and scaled. Understanding these vulnerabilities is essential for developing more robust protection systems against increasingly sophisticated prompt injection attacks.

One-Shot is Enough: Consolidating Multi-Turn Attacks into Efficient Single-Turn Prompts for LLMs