Skip to main content
LLM Security Database
Skip to research search
Last analyzed Sep 9, 2026

Language Model Security Database

985 research findings · 1119 evaluated models

Filtered research findings

798 entries

Matches every word across titles, descriptions, sources, affected systems, and models.

Unlearning Relearning Attack

Published Analyzed Source: arXiv

Standard Large Language Model (LLM) unlearning techniques, specifically Negative Preference Optimization (NPO), Gradient Difference (GradDiff), and Representation Misdirection for Unlearning (RMU), fail to sufficiently flatten the loss landscape surrounding the "forgotten" weights. This sharp loss landscape allows for a "Relearning Attack," wherein an attacker can fully restore the unlearned capabilities (such as hazardous knowledge, sensitive data, or copyrighted material) by performing…

Towards llm unlearning resilient to relearning attacks: A sharpness-aware minimization perspective and beyond
Evaluated models: Llama 2 7B, Llama 3 8B

Simple Interaction Jailbreaks

Published Analyzed Source: arXiv

Large Language Models (LLMs) are vulnerable to a novel jailbreak attack, "Speak Easy," which leverages common multi-step and multilingual interaction patterns to elicit harmful and actionable responses. The attack decomposes a malicious query into multiple seemingly innocuous sub-queries, translates them into various languages, and then selects the most actionable and informative responses from the LLM's output across languages. This bypasses existing safety mechanisms more effectively than…

Speak Easy: Eliciting Harmful Jailbreaks from LLMs with Simple Interactions
Evaluated models: GPT-4o, Llama 3.1 8B Instruct, Llama 3.3 70B Instruct+1 more
  • GPT-4o
  • Llama 3.1 8B Instruct
  • Llama 3.3 70B Instruct
  • Qwen 2 72B Instruct

Distilled Jailbreak Prompt Generator

Published Analyzed Source: arXiv

The Knowledge-Distilled Attacker (KDA) model, when used to generate prompts for large language models (LLMs), can bypass LLM safety mechanisms resulting in the generation of harmful, inappropriate, or misaligned content. KDA's effectiveness stems from its ability to generate diverse and coherent attack prompts efficiently, surpassing existing methods in attack success rate and speed. The vulnerability lies in the LLMs' insufficient defenses against the diverse prompt generation strategies…

KDA: A Knowledge-Distilled Attacker for Generating Diverse Prompts to Jailbreak LLMs
Evaluated models: Claude 2.1, GPT-3.5 Turbo, GPT-4+8 more
  • Claude 2.1
  • GPT-3.5 Turbo
  • GPT-4
  • GPT-4 Turbo
  • Llama 2 13B Chat
  • Llama 2 7B Chat
  • Mistral 7B
  • Qwen 14B Chat
  • Qwen 7B Chat
  • Vicuna 13B
  • Vicuna 7B

Intent Flattening Jailbreak

Published Analyzed Source: arXiv

A vulnerability exists in the safety alignment mechanisms of Large Language Models (LLMs) related to the model's intent perception capabilities. The specific attack vector, termed "Perceived-importance Flatten" (PiF), circumvents safety guardrails by modifying neutral-intent tokens within a malicious prompt using synonym replacement. Unlike traditional jailbreak attacks that rely on appending lengthy, high-perplexity adversarial suffixes (which suffer from distributional dependency and often…

Understanding and Enhancing the Transferability of Jailbreaking Attacks
Evaluated models: Llama 2 13B Chat, Llama 3.1 8B Instruct, Mistral 7B Instruct+5 more
  • Llama 2 13B Chat
  • Llama 3.1 8B Instruct
  • Mistral 7B Instruct
  • Vicuna 13B v1.5
  • GPT-4 0613
  • o1-preview
  • Claude 3.5 Sonnet
  • Gemini 1.5 Flash

AP-Test Guardrail Identification

Published Analyzed Source: arXiv

This vulnerability allows attackers to identify the presence and location (input or output stage) of specific guardrails implemented in Large Language Models (LLMs) by using carefully crafted adversarial prompts. The attack, termed AP-Test, leverages a tailored loss function to optimize these prompts, maximizing the likelihood of triggering a specific guardrail while minimizing triggering others. Successful identification provides attackers with valuable information to design more effective…

Peering Behind the Shield: Guardrail Identification in Large Language Models
Evaluated models: Aegis Defensive, Aegis Permissive, GPT-4o+8 more
  • Aegis Defensive
  • Aegis Permissive
  • GPT-4o
  • Llama Guard
  • Llama Guard 2
  • Llama Guard 3
  • Perspective
  • ShieldGemma 2B
  • ShieldGemma 9B
  • ShieldGemma 27B
  • WildGuard

Adversarial LLM Jailbreak

Published Analyzed Source: arXiv

A vulnerability in Large Language Models (LLMs) allows adversarial reasoning attacks to bypass safety mechanisms and elicit harmful responses. The vulnerability stems from the insufficient robustness of existing LLM safety measures against iterative prompt refinement guided by a loss function that measures the LLM's proximity to generating a target harmful response. This allows an attacker to effectively navigate the prompt space, even against adversarially trained models, resulting in…

Adversarial Reasoning at Jailbreaking Time
Evaluated models: Claude 3.5 Sonnet, Cygnet, Gemini 1.5 Pro+8 more
  • Claude 3.5 Sonnet
  • Cygnet
  • Gemini 1.5 Pro
  • GPT-4
  • Llama 2 7B
  • Llama 3 8B
  • Llama 3.1 405B
  • Mixtral 8x7B
  • o1-preview
  • R2D2
  • Vicuna 13B v1.5

Inherited GPT Policy Violations

Published Analyzed Source: arXiv

A policy compliance vulnerability exists in the OpenAI GPT Store ecosystem affecting Custom GPTs. The vulnerability stems from the inheritance of safety alignment weaknesses from foundational models (GPT-4 and GPT-4o) and the insufficient enforcement of usage policies during the customization and review process. Custom GPTs can be trivially manipulated to violate safety guidelines—specifically regarding Cybersecurity (malware generation), Academic Integrity (ghostwriting), and Romantic…

Towards Safer Chatbots: A Framework for Policy Compliance Evaluation of Custom GPTs
Evaluated models: GPT-4, GPT-4o

Topic-Flip RAG Poisoning

Published Analyzed Source: arXiv

Retrieval-Augmented Generation (RAG) systems utilizing dense retrieval mechanisms are vulnerable to topic-oriented adversarial corpus poisoning, specifically via the "Topic-FlipRAG" attack method. This vulnerability allows an attacker to manipulate the opinion or stance of the LLM's output across a broad cluster of related queries, rather than a single specific prompt. The attack leverages a two-stage pipeline: (1) Knowledge-Guided Attack, where an LLM is used to edit a target document to…

Topic-fliprag: Topic-orientated adversarial opinion manipulation attacks to retrieval-augmented generation models
Evaluated models: GPT-4o, Llama 3.1 8B, Qwen 2.5 7B+1 more
  • GPT-4o
  • Llama 3.1 8B
  • Qwen 2.5 7B
  • o4-mini

Weight Tampering Unlearning Bypass

Published Analyzed Source: arXiv

State-of-the-art machine unlearning and safety fine-tuning methods for Large Language Models (LLMs) fail to robustly remove hazardous capabilities or refusal mechanisms from model weights. While these methods suppress model outputs during standard input-output interactions, the underlying capabilities remain latent in the parameter space. An attacker with access to model weights (e.g., via open releases or leaked weights) can restore "unlearned" knowledge (such as dual-use biology hazards) or…

Model tampering attacks enable more rigorous evaluations of llm capabilities
Evaluated models: Llama 3 8B

Multimodal Flanking Jailbreak

Published Analyzed Source: arXiv

A novel "Flanking Attack" exploits the vulnerability of multimodal LLMs (e.g., Google Gemini) to bypass content moderation filters by embedding adversarial prompts within a sequence of benign prompts. The attack leverages the LLM's processing of both audio and text, obfuscating harmful requests through contextualization and layering, thereby yielding policy-violating responses.

From Compliance to Exploitation: Jailbreak Prompt Attacks on Multimodal LLMs
Evaluated models: Not reported

Research methodology

Entries summarize publicly available primary-source security research. Model names reflect only systems explicitly evaluated by the cited paper, and measurements are research-reported unless independent verification is stated.

Discovery and publication are separate: new papers enter a candidate queue, then require human review and explicit approval before publication. Updates follow that review process, not a guaranteed daily publishing schedule. “Last analyzed” is the newest analysis date among published entries; it does not mean that all recent research has been covered.