Pith. sign in

REVIEW 14 cited by

Visual Adversarial Examples Jailbreak Aligned Large Language Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2306.13213 v2 pith:ZBRLIKUZ submitted 2023-06-22 cs.CR cs.CLcs.LG

classification cs.CRcs.CLcs.LG
keywords adversarialvisualllmsmodelsalignedlanguagealignmentattack
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Recently, there has been a surge of interest in integrating vision into Large Language Models (LLMs), exemplified by Visual Language Models (VLMs) such as Flamingo and GPT-4. This paper sheds light on the security and safety implications of this trend. First, we underscore that the continuous and high-dimensional nature of the visual input makes it a weak link against adversarial attacks, representing an expanded attack surface of vision-integrated LLMs. Second, we highlight that the versatility of LLMs also presents visual attackers with a wider array of achievable adversarial objectives, extending the implications of security failures beyond mere misclassification. As an illustration, we present a case study in which we exploit visual adversarial examples to circumvent the safety guardrail of aligned LLMs with integrated vision. Intriguingly, we discover that a single visual adversarial example can universally jailbreak an aligned LLM, compelling it to heed a wide range of harmful instructions that it otherwise would not) and generate harmful content that transcends the narrow scope of a `few-shot' derogatory corpus initially employed to optimize the adversarial example. Our study underscores the escalating adversarial risks associated with the pursuit of multimodality. Our findings also connect the long-studied adversarial vulnerabilities of neural networks to the nascent field of AI alignment. The presented attack suggests a fundamental adversarial challenge for AI alignment, especially in light of the emerging trend toward multimodality in frontier foundation models.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 14 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Securing Multimodal AI through Internal Information Decomposition

    cs.AI 2026-05 conditional novelty 6.0 of 10

    A one-class detector using first-token distributional consistency between text-only, vision-only, and joint predictions reduces multimodal jailbreak attack success rates to below 15% with ~2.4% utility loss.

  2. SoK: Systematizing LLM Prompt Security: Taxonomies, Datasets, and Unified Evaluation of Attacks and Defenses

    cs.CR 2025-10 conditional novelty 6.0 of 10

    A systemization of LLM jailbreak security that adds linked taxonomies, an evaluation platform, and JailbreakDB, while its main attack–defense comparison results remain deferred.

  3. VISOR++: Universal Visual Inputs based Steering for Large Vision Language Models

    cs.CV 2025-09 conditional novelty 6.0 of 10

    A single adversarially optimized image can reproduce activation-steering behavior in multiple VLMs and partially transfer to unseen models.

  4. Activation Steering Meets Preference Optimization: Defense Against Jailbreaks in Vision Language Models

    cs.CV 2025-08 reject novelty 6.0 of 10

    A proposed VLM defense, SPO-VLM, combines activation steering with sequence-level preference optimization and claims lower jailbreak ASR and toxicity than ASTRA while retaining visual understanding.

  5. Bridging the Gap in Vision Language Models in Identifying Unsafe Concepts Across Modalities

    cs.CR 2025-07 conditional novelty 6.0 of 10

    Vision-language models consistently recognize unsafe content better from text than from images, and a simplified reinforcement learning fine-tune narrows that gap.

  6. AMIA: Automatic Masking and Joint Intention Analysis Makes LVLMs Robust Jailbreak Defenders

    cs.CV 2025-05 conditional novelty 6.0 of 10

    AMIA is an inference-only jailbreak defense that masks text-irrelevant image patches and prompts single-pass intention analysis, lifting average defense success on LVLMs from 52.4% to 81.7%.

  7. Bootstrapping LLM Robustness for VLM Safety via Reducing the Pretraining Modality Gap

    cs.AI 2025-05 conditional novelty 6.0 of 10

    Reducing the modality gap between image and text embeddings during LVLM pretraining reduces unsafe response rates by up to 16.3% across models and benchmarks.

  8. Effective Black-Box Multi-Faceted Attacks Breach Vision Large Language Model Guardrails

    cs.CV 2025-02 conditional novelty 6.0 of 10

    A black-box attack combining visual prompt injection, a contrasting-responses jailbreak, and a moderator-fooling suffix achieves 61.56% average success on eight commercial VLLMs.

  9. Layer-Wise Perturbations via Sparse Autoencoders for Adversarial Text Generation

    cs.CL 2025-08 reject novelty 5.0 of 10

    Sparse autoencoder activation perturbation (SFPF) applied on top of existing jailbreak prompts raises attack success rate on Qwen3-32B, but with no defense evaluation and weak reproducibility.

  10. Seeing the Threat: Vulnerabilities in Vision-Language Models to Adversarial Attack

    cs.CL 2025-05 conditional novelty 5.0 of 10

    A two-stage evaluation framework and token-projection analysis show that LVLMs encode harmful semantic cues from images even without OCR, while remaining vulnerable to cross-modal attacks.

  11. VLM-Guard: Safeguarding Vision-Language Models via Fulfilling Safety Alignment Gap

    cs.CR 2025-02 conditional novelty 4.0 of 10

    VLM-Guard steers VLM hidden states along the safety steering direction of the aligned LLM component, cutting attack success rate on LLaVA-1.5-7b from 15-72% to 4-7% across three settings.

  12. Universal Adversarial Attack on Aligned Multimodal LLMs

    cs.AI 2025-02 conditional novelty 4.0 of 10

    A single optimized image, trained through the vision and language modules, makes aligned multimodal LLMs produce dangerous responses across diverse prompts and some models.

  13. Pushing the Limits of Safety: A Technical Report on the ATLAS Challenge 2025

    cs.CR 2025-06 conditional novelty 3.0 of 10

    The ATLAS 2025 competition demonstrates that vision-language models remain highly vulnerable to flowchart-based and cross-modal jailbreak attacks, with top scores exceeding 93%.

  14. From Hallucinations to Jailbreaks: Rethinking the Vulnerability of Large Foundation Models

    cs.CV 2025-05 reject novelty 3.0 of 10

    The paper argues hallucinations and jailbreaks share the same optimization dynamics and shows that defenses for one also reduce the other, but the theoretical support is largely circular.

Pith tools