Language Model Security Database

LLM agents equipped with tool-use, persistent memory, and environmental interaction capabilities are vulnerable to long-horizon attacks. Attackers can bypass single-turn safety guardrails by exploiting the temporal dimension of multi-turn interactions to incrementally steer agent behavior. The vulnerability manifests because the agent's safety mechanisms perform localized, single-step evaluations but fail to maintain semantic safety across extended interaction trajectories. This enables…

Large Language Models (LLMs) supporting function calling (tool use) are vulnerable to a jailbreak attack known as iMIST (interactive Multi-step Progressive Tool-disguised Jailbreak Attack). The vulnerability stems from a disparity in alignment training: while models are heavily aligned to refuse harmful natural language generation, they lack sufficient alignment regarding the generation of harmful content within structured data (JSON) used for tool parameters.

A vulnerability exists in Large Language Model (LLM) deployments and multi-agent systems where an autonomous attacker agent can systematically extract hidden system prompts through self-evolving interaction strategies. The vulnerability leverages a "JustAsk" framework which utilizes Upper Confidence Bound (UCB) exploration to dynamically select and refine attack vectors from a hierarchical taxonomy of 14 atomic skills (e.g., structural formatting, authority appeals) and 14 multi-turn…

Commercial Multimodal Large Language Model (MLLM) integrated systems are vulnerable to a "Dual Steganography" jailbreak paradigm (referred to as Odysseus). The vulnerability arises from the reliance of safety filters on the assumption that malicious content must be explicitly visible in the input or output modalities (text or image). Attackers can bypass these filters by encoding malicious queries into binary matrices and embedding them into benign-looking images using steganographic encoders…

Reasoning-specialized Large Language Models (LLMs) that utilize Chain-of-Thought (CoT) processes are vulnerable to reasoning-exploitation jailbreaks. Attackers can bypass standard safety alignments (such as RLHF) by using adaptive multi-turn interactions or semantic transformations to induce the model to generate intermediate reasoning steps that "rationalize" or "contextualize" a harmful request. Because current alignment techniques often fail to scale linearly with reasoning depth, forcing…

A safety bypass vulnerability exists in Large Language Models (LLMs) and Multi-Modal Models (LVLMs) that expose a "Prefix Completion" parameter in their inference APIs (e.g., echo or prefix parameters). The vulnerability exploits the model's autoregressive nature, where the model prioritizes local coherence with a user-supplied output prefix over global safety alignment (RLHF) or system instructions. By supplying a prefix that establishes a hostile narrative or intent (e.g., "My actions are as…

A vulnerability exists in the tool selection mechanisms of Large Language Model (LLM) agents, identified as the "Attractive Metadata Attack" (AMA). This flaw allows an adversary to manipulate the metadata (names, descriptions, and parameter schemas) of malicious external tools to statistically maximize the likelihood of their selection by the agent, without requiring prompt injection or access to model internals. The vulnerability exploits the agent’s semantic scoring function used to map user…

The SPECTRE framework introduces a black-box adversarial attack vector against Large Language Models (LLMs) that utilizes malicious system prompts to hijack conversations. Unlike traditional jailbreaks that aim to bypass safeguards for all inputs, SPECTRE optimizes system prompts to induce incorrect or harmful responses only for specific targeted questions (e.g., "Are COVID vaccines safe?", "Who should I vote for?"), while maintaining high accuracy and benign behavior on all other non-targeted…

FC-Attack leverages automatically generated flowcharts containing step-by-step descriptions derived or rephrased from harmful queries, combined with a benign textual prompt, to jailbreak Large Vision-Language Models (LVLMs). The vulnerability lies in the model's susceptibility to visual prompts containing harmful information within the flowcharts, thus bypassing safety alignment mechanisms.

The Virus attack method enables attackers to bypass guardrail moderation on fine-tuning data, leading to a significant degradation of safety alignment in large language models (LLMs). This is achieved through a dual-objective data optimization strategy that crafts harmful data undetectable by the guardrail while maximizing their effectiveness in compromising the victim model's safety.