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

337 entries

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

LLM Judge Manipulation

Published Analyzed Source: arXiv

Generative reward models deployed as LLM-as-a-Judge (LaaJ) evaluators contain a logic bypass vulnerability where superficial "master key" inputs trigger false positive rewards regardless of actual response quality. Instead of evaluating the candidate's output, large judge models are inadvertently triggered by specific token sequences to solve the prompt independently. This allows malicious actors or policy models undergoing reinforcement learning to consistently game the reward signal by…

Security in LLM-as-a-Judge: A Comprehensive SoK
Evaluated models: GPT-4o, o1, Qwen 2.5 72B Instruct+1 more
  • GPT-4o
  • o1
  • Qwen 2.5 72B Instruct
  • Llama 3 70B Instruct

Agentic Robot Instruction Attack

Published Analyzed Source: arXiv

Vision-Language-Action (VLA) models are vulnerable to targeted, low-budget textual perturbations in their natural-language instruction inputs, which can maliciously alter sequential decision-making and downstream physical robotic behavior. Because VLA policies tightly couple language, perception, and control, bounded edits—such as character-level typos, token attribute swaps, or prompt-level uncertainty clauses—propagate through the model's execution trajectory. This allows a black-box…

SABER: A Stealthy Agentic Black-Box Attack Framework for Vision-Language-Action Models
Evaluated models: Not reported

RAG Arbitrary Query Poisoning

Published Analyzed Source: arXiv

A compound vulnerability in Retrieval-Augmented Generation (RAG) systems allows attackers to deterministically hijack model outputs for arbitrary user queries without prior knowledge of the user's input. The vulnerability, identified as PIDP-Attack, requires a dual-vector exploitation: database poisoning and query-path prompt injection. First, the attacker injects a small number of poisoned passages into the RAG database, each starting with an attacker-chosen "target question" followed by a…

PIDP-Attack: Combining Prompt Injection with Database Poisoning Attacks on Retrieval-Augmented Generation Systems
Evaluated models: Llama 3.1 8B, Qwen 2 7B, Qwen 2.5 7B

Autoresearch Adversarial Algorithm Discovery

Published Analyzed Source: arXiv

GPT-OSS-Safeguard-20B and Meta-SecAlign (70B/8B) are vulnerable to white-box adversarial attacks generated by automated algorithmic recombination (specifically the claude_v63, claude_v82, and claude_v53-oss optimizers). These algorithms significantly outperform standard discrete optimization methods (like GCG) by integrating continuous optimization (ADC) with LayerNorm gradient scaling (LSGM), or by merging momentum-smoothed gradients with directional perturbation candidate selection (DPTO)…

Claudini: Autoresearch Discovers State-of-the-Art Adversarial Attack Algorithms for LLMs
Evaluated models: Llama 2 7B, Llama 3 8B, Qwen 2.5 7B+2 more
  • Llama 2 7B
  • Llama 3 8B
  • Qwen 2.5 7B
  • Gemma 2 2B
  • Gemma 7B

MCP Tool-Metadata Prompt Injection in Coding Agents

Published Analyzed Source: arXiv

The paper presents a controlled, reproducible evaluation in which untrusted MCP tool descriptions contain hidden instructions that attempt to induce coding agents to access files, invoke other tools, create deceptive links, or execute remote scripts. The authors report that several clients accepted poisoned metadata without adequate validation or parameter visibility; Cursor 1.6.45 was reported unsafe across all four tested vectors. These are paper-reported results from isolated local tests…

Are AI-assisted Development Tools Immune to Prompt Injection?
Evaluated models: Claude Sonnet 4.5, Grok Code Fast 1, Gemini 2.5 Pro+1 more
  • Claude Sonnet 4.5
  • Grok Code Fast 1
  • Gemini 2.5 Pro
  • Claude Opus 4

Profit-Driven Agent Exploitation

Published Analyzed Source: arXiv

LLM-based autonomous agents deployed in multi-turn, structured environments are vulnerable to adaptive, profit-driven semantic exploitation. Rather than utilizing traditional malformed prompt injections or jailbreaks, an attacker can leverage valid interaction channels to execute social engineering, protocol spoofing, and authority impersonation tactics. By strategically shaping the environment's context—such as feigning technical constraints, fabricating evaluation harnesses, or manipulating…

Profit is the Red Team: Stress-Testing Agents in Strategic Economic Interactions
Evaluated models: Not reported

Stage-Sequential Agent Escalation

Published Analyzed Source: arXiv

Agentic Large Language Model (LLM) systems utilizing persistent memory, Retrieval-Augmented Generation (RAG) pipelines, and external tool connectors are vulnerable to Logic-layer Prompt Control Injection (LPCI). An attacker can inject obfuscated (e.g., encoded, structurally nested, or semantically reframed) payloads into external memory stores or RAG documents. These payloads bypass conventional inference-time plaintext content filters, persist across session boundaries, and remain dormant…

LAAF: Logic-layer Automated Attack Framework A Systematic Red-Teaming Methodology for LPCI Vulnerabilities in Agentic Large Language Model Systems
Evaluated models: GPT-4o Mini, Claude 3 Haiku, Llama 3.1 70B Instruct+2 more
  • GPT-4o Mini
  • Claude 3 Haiku
  • Llama 3.1 70B Instruct
  • Gemini 2.0 Flash
  • Mixtral 8x7B Instruct

Contextual Priority Hacking

Published Analyzed Source: arXiv

Large Language Models (LLMs) are vulnerable to a jailbreak technique termed "Priority Hacking." Adversaries can bypass safety alignments by exploiting the model's internal priority graph, where certain abstract values (e.g., justice, public health) implicitly outweigh general safety restrictions within specific contexts. By crafting a deceptive prompt that frames a malicious request as a necessary action in service of a higher-priority benign value, attackers engineer a value conflict. The…

Are Dilemmas and Conflicts in LLM Alignment Solvable? A View from Priority Graph
Evaluated models: Not reported

Inaudible Ultrasonic LLM Jailbreak

Published Analyzed Source: arXiv

Speech-driven Large Language Models (LLMs) and end-to-end Large Audio-Language Models (LALMs) are vulnerable to inaudible near-ultrasonic prompt injections, a framework dubbed Sirens' Whisper (SWhisper). By exploiting the non-linear response of commodity microphones, attackers can encode structured, phonetically optimized adversarial prompts into the 17–22 kHz near-ultrasonic band. Using regularized channel-inversion pre-compensation, the attacker shapes the waveform to account for microphone…

Sirens' Whisper: Inaudible Near-Ultrasonic Jailbreaks of Speech-Driven LLMs
Evaluated models: GLM-4 Voice, Qwen Omni Turbo, Llama 3.1 8B Instruct+5 more
  • GLM-4 Voice
  • Qwen Omni Turbo
  • Llama 3.1 8B Instruct
  • Gemma 3 4B
  • Qwen 2.5 7B Instruct
  • Mistral 7B Instruct v0.3
  • GLM-4 Air 250414
  • Grok 4

Autonomous Agent Tool RCE

Published Analyzed Source: arXiv

The OpenClaw autonomous agent framework lacks execution sandboxing, running agents directly on the host machine with the disk and system privileges of the host user. This architecture allows attackers to achieve Remote Code Execution (RCE) and arbitrary data exfiltration via Indirect Prompt Injection. By embedding malicious instructions within external data sources (e.g., scraped web pages or uploaded documents), an attacker can hijack the agent's planning capabilities to sequentially chain…

Uncovering Security Threats and Architecting Defenses in Autonomous Agents: A Case Study of OpenClaw
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.