AI Threat Alert
When LLM Defenses Backfire: Characterizing Safety, Performance, and Cost Trade-offs
Jailbreak defenses are essential for protecting large language models (LLMs
AgentLens: Interpretable Safety Steering via Mechanistic Subspaces for Multi-Turn Coding Agent
behavioral control during execution. Meanwhile, recent mechanistic interpretability methods for LLM safety are mostly confined to single-turn or jailbreak-style QA settings, limiting their ability to capture the evolving
KG-DF: A Black-box Defense Framework against Jailbreak Attacks Based on Knowledge Graphs
show that our framework enhances defense performance against various jailbreak attack methods, while also improving the response quality of the LLM in general QA scenarios by incorporating domain-general knowledge
Invasive Context Engineering to Control Large Language Models
good results, LLMs remain susceptible to abuse, and jailbreak probability increases with context length. There is a need for robust LLM security guarantees in long-context situations. We propose control
Multi-Turn Jailbreaking of Aligned LLMs via Lexical Anchor Tree Search
Tree Search (), addressing these limitations through an attacker-LLM-free method that operates purely via lexical anchor injection. LATS reformulates jailbreaking as a breadth-first tree search over multi-turn
Inspect India Evals: An Open Benchmarking Framework for Evaluating Large Language Models in the Indian Linguistic and Cultural Context
using harmful prompts in Indian languages, a multi-turn jailbreak resistance test, and an Indian cultural knowledge benchmark scored using LLM-as-judge rubrics. In this study, we tested five
Silent Alarm: A J-Space Protocol for Comparing Danger Recognition Across Models and Quantization Levels
Jailbreak-robustness research typically evaluates safety through generated responses using an LLM-as-judge approach. Such evaluations, however, are sensitive to the benchmark's grading procedure and capture only observed
Analysing the Safety Pitfalls of Steering Vectors
that steering vectors consistently influence the success rate of jailbreak attacks, with stronger amplification under simple template-based attacks. Across LLM families and sizes, steering the model in specific directions
Untargeted Jailbreak Attack
Existing gradient-based jailbreak attacks on Large Language Models (LLMs) typically optimize adversarial suffixes to align the LLM output with predefined target responses. However, restricting the objective as inducing fixed
ToxScreen: Detecting Whether an LLM Has Been Poisoned
between attack behaviors and the weights of an LLM. We find a phenomenon whereby backdoors operate via different mechanistic strategies than jailbreaks, allowing defenders to filter jailbreaks. Finally, no method
Geometric Configurations of Perturbed Jailbreak Prompts
Perturbation techniques that turn unsuccessful jailbreak prompts into successful ones are continuously evolving, constituting a major security threat to LLM safety. In this paper, we investigate the internal representations
EquaCode: A Multi-Strategy Jailbreak Approach for Large Language Models via Equation Solving and Code Completion
significant concern, as they are still susceptible to jailbreak attacks aimed at eliciting inappropriate or harmful responses. However, existing jailbreak attacks mainly operate at the natural language level and rely
The Silicon Psyche: Anthropomorphic Vulnerabilities in Large Language Models
adversarial scenarios targeting LLM decision-making. Our preliminary hypothesis testing across seven major LLM families reveals a disturbing pattern: while models demonstrate robust defenses against traditional jailbreaks, they exhibit critical
RunawayEvil: Jailbreaking the Image-to-Video Generative Models
strategies through reinforcement learning-driven strategy customization and LLM-based strategy exploration; (2) Multimodal Tactical Planning Unit that generates coordinated text jailbreak instructions and image tampering guidelines based
A Causal Perspective for Enhancing Jailbreak Attack and Defense
causal approaches. Our results suggest that analyzing jailbreak features from a causal perspective is an effective and interpretable approach for improving LLM reliability. Our code is available at https://github.com
Multi-Agent Framework for Threat Mitigation and Resilience in AI-Based Systems
stages. Results: We identify unreported threats including commercial LLM API model stealing, parameter memorization leakage, and preference-guided text-only jailbreaks. Dominant TTPs include MASTERKEY-style jailbreaking, federated poisoning, diffusion
Benchmark of Benchmarks: Unpacking Influence and Code Repository Quality in LLM Safety Benchmarks
based on both automated and human assessment) on LLM safety benchmarks, analyzing 31 benchmarks and 382 non-benchmarks across prompt injection, jailbreak, and hallucination. We find that benchmark papers show
Countermind: A Multi-Layered Security Architecture for Large Language Models
security of Large Language Model (LLM) applications is fundamentally challenged by "form-first" attacks like prompt injection and jailbreaking, where malicious instructions are embedded within user inputs. Conventional defenses, which
Assessing Automated Prompt Injection Attacks in Agentic Environments
prompt injection poses a critical threat to LLM agents that interact with untrusted external data, yet automated attack methods--proven effective for jailbreaking--remain underexplored in realistic agentic settings
Jailbreak-Zero: A Path to Pareto Optimal Red Teaming for Large Language Models
This paper introduces Jailbreak-Zero, a novel red teaming methodology that shifts the paradigm of Large Language Model (LLM) safety evaluation from a constrained example-based approach to a more