Black-box Vulnerabilities

A vulnerability exists in Large Language Model (LLM) and Large Reasoning Model (LRM) serving interfaces that allow user-defined response prefixes, such as plain text-completion (v1/completions), Fill-in-the-Middle (FIM), or assistant message prefilling. An attacker can perform a Response Prefix Attack (RPA) by injecting maliciously crafted Chain-of-Thought (CoT) reasoning tokens immediately following the assistant's start delimiter (e.g., <|im_start|>assistant). Because these tokens are placed…

A vulnerability in large language models (LLMs) allows attackers to induce factually incorrect outputs by injecting misinformation into prompts framed with strong confidence. By using authoritative phrasing (e.g., "As we know..."), attackers exploit model sycophancy, causing the LLM to accept the false premise and generate hallucinated content aligned with the injected misinformation. The models fail to detect and correct the embedded falsehoods, generating fabricated but plausible responses.

Multi-modal Large Language Models (MLLMs) capable of processing interleaved image-text sequences are vulnerable to a universal adversarial perturbation (UAP) attack known as LAMP. This vulnerability allows an attacker to generate a single, noise-based perturbation pattern using a surrogate model (e.g., Mantis-CLIP) that transfers effectively to black-box target models. The attack leverages two novel loss functions during perturbation learning: a "contagious" objective that manipulates…

A vulnerability exists in frontier Large Language Models (LLMs) where in-context information (e.g., provided via Retrieval-Augmented Generation) completely overrides parametric safety guardrails when processing counterfactual or adversarial medical evidence. When a prompt contains fabricated clinical context asserting the medical efficacy of toxic substances, illicit drugs, or nonsensical items, the LLM suppresses its internal knowledge of the substance's toxicity. Internal representation…

Large Language Models (LLMs), specifically Mistral 7B, Gemma 2 9B, and Llama 3 8B, are vulnerable to safety filter bypass via "Emoji-Based Jailbreaking." This adversarial prompt engineering technique exploits the model's tokenization and internal representation of Unicode emoji characters. By utilizing "emoji stuffing" (inserting emojis between textual tokens) or "emoji chaining" (using sequences of emojis as semantic proxies for sensitive terms), attackers can evade keyword-based safety…

A vulnerability in the prompt-side safety filters of GPT-based Text-to-Image (T2I) systems allows attackers to bypass restrictions on Politically Sensitive Content (PSC). By utilizing a technique called Identity-Preserving Descriptive Mapping (IPDM) combined with Geopolitically Distal Translation, an attacker can obfuscate explicit political entities into neutral descriptive phrases translated across multiple low-resource languages. This induces semantic fragmentation, preventing the safety…

A vulnerability exists in aligned Large Language Models (LLMs) where inducing "drunk language" behavior—simulating the text of an intoxicated human—bypasses safety guardrails and contextual privacy protections. Attackers can exploit this anthropomorphic flaw through inference-time persona prompting or lightweight post-training (causal fine-tuning or reinforcement learning on drunk text corpora). By forcing the model to adopt a stylistic and semantic framework associated with impaired human…

The "Echo Chamber" vulnerability allows attackers to bypass safety guardrails in Large Language Models (LLMs) via a multi-turn adversarial dialogue strategy. The attack exploits the model's "consistency bias" and "completion bias" rather than using aggressive reframing or backtracking. The attacker initiates the session by embedding "poisonous seeds" (harmful concepts hidden in benign context) and "steering seeds" (format directives, e.g., "write a manual") into an ostensibly innocuous prompt…

A "Gamified Adversarial Multimodal Breakout via Instructional Traps" (GAMBIT) vulnerability exists in the safety alignment mechanisms of Multimodal Large Language Models (MLLMs), specifically those employing Chain-of-Thought (CoT) reasoning. The vulnerability exploits the finite cognitive resource budget of the model by inducing "cognitive overload" through a high-stakes, gamified context. The attack functions by decomposing a harmful query into a visual puzzle (e.g., a shuffled grid of image…

Large Language Models (LLMs) are vulnerable to a gray-box adversarial attack method known as RAILS (RAndom Iterative Local Search). This vulnerability allows an attacker with access to model output logits (but without access to gradients or weights) to optimize discrete adversarial suffixes that bypass safety alignment. The attack employs a random local search guided by a hybrid loss function combining Teacher-Forcing and a novel Auto-Regressive loss that enforces exact target prefix matching…