Safety Vulnerabilities

Large Language Models (LLMs) are vulnerable to jailbreak attacks using language games, which manipulate input prompts through structured linguistic alterations (e.g., Ubbi Dubbi, custom letter insertion rules) to bypass safety mechanisms. These games obfuscate malicious intent while maintaining human readability, causing LLMs to generate unsafe content.

Large Language Models (LLMs) are vulnerable to multi-round jailbreak attacks which leverage a heuristic search process to progressively elicit harmful content. The attack decomposes a harmful query into multiple, seemingly innocuous sub-queries, iteratively refining the prompts based on the LLM's responses and employing psychological strategies to bypass safety mechanisms. This allows for the circumvention of single-round detection methods and elicitation of responses containing prohibited…

Large Language Models (LLMs) are vulnerable to multi-step "moralized" jailbreak prompts that bypass their safety guardrails. These prompts, while appearing ethical individually, cumulatively create a context that elicits verbally aggressive and harmful content generation. The attack leverages the LLMs' inability to fully understand the cumulative context and intent across multiple prompts.

Large language models (LLMs) are vulnerable to jailbreak attacks exploiting nonlinear features within prompt encodings. These features, not detectable by linear methods, allow adversaries to reliably elicit harmful outputs despite safety training. Different attack methods leverage distinct nonlinear features, limiting the transferability of detection and mitigation techniques.

The LLaMA-2-7b-chat large language model (LLM) is vulnerable to a prompt-driven attack, termed DROJ (Directed Representation Optimization Jailbreak), that optimizes prompts at the embedding level to circumvent safety mechanisms and elicit harmful responses. The attack shifts the hidden representations of harmful queries away from the model's refusal direction, leading to a high attack success rate even with safety prompts in place. While the model may not refuse, responses may be repetitive…

LLMStinger uses a reinforcement-learning loop to fine-tune an attacker model that generates adversarial suffixes for jailbreak prompts. Because the approach does not require white-box access to the target, it can adapt existing attacks for both open- and closed-source assistants and bypass otherwise effective refusal behavior.

A novel SQL Injection Jailbreak (SIJ) vulnerability allows attackers to bypass safety mechanisms in Large Language Models (LLMs) by manipulating the structure of input prompts. The attack leverages the model's processing of system prompts, user prefixes, user prompts, and assistant prefixes to effectively "comment out" the expected response prefix and inject harmful instructions, causing the LLM to generate unsafe content. This vulnerability exploits the external properties of the LLM…

Large Language Models (LLMs) are vulnerable to "SequentialBreak," a jailbreak attack where embedding a harmful prompt within a chain of benign prompts in a single query can bypass LLM safety features. The LLM's attention mechanism prioritizes the benign prompts, allowing the harmful prompt to be processed without triggering safety mitigations.

Large Language Models (LLMs) employing safety alignment mechanisms are vulnerable to a bypass attack using simple, stochastic random augmentations of input prompts. The attack leverages the inherent brittleness of safety alignment to minor, randomly introduced modifications in the input, causing the LLM to generate unsafe outputs despite its safety training. Character-level augmentations prove significantly more effective than string insertions.

Large Language Models (LLMs) are vulnerable to a novel prompt injection attack using universal and context-independent triggers. These triggers, once discovered for a specific LLM, allow precise control over the model's output regardless of the prompt context or desired output content, enabling adversaries to force the generation of arbitrary text. The attack utilizes a gradient-based optimization technique to discover these triggers.