Learning To See But Forgetting To Follow: Visual Instruction Tuning Makes LLMs More Prone To Jailbreak Attacks

TL;DR

This paper investigates how visual instruction tuning in vision-language models increases their vulnerability to jailbreaking attacks, highlighting safety concerns and proposing evaluation strategies to mitigate these risks.

Contribution

It reveals that visual instruction tuning causes forgetting of safety guardrails in LLMs, making VLMs more prone to jailbreaking compared to their LLM backbones.

Findings

VLMs are more susceptible to jailbreaking than LLMs.

Visual instruction tuning causes safety guardrail forgetting.

Recommendations for evaluation strategies to improve safety.

Abstract

Augmenting Large Language Models (LLMs) with image-understanding capabilities has resulted in a boom of high-performing Vision-Language models (VLMs). While studying the alignment of LLMs to human values has received widespread attention, the safety of VLMs has not received the same attention. In this paper, we explore the impact of jailbreaking on three state-of-the-art VLMs, each using a distinct modeling approach. By comparing each VLM to their respective LLM backbone, we find that each VLM is more susceptible to jailbreaking. We consider this as an undesirable outcome from visual instruction-tuning, which imposes a forgetting effect on an LLM's safety guardrails. Therefore, we provide recommendations for future work based on evaluation strategies that aim to highlight the weaknesses of a VLM, as well as take safety measures into account during visual instruction tuning.