LLM Jailbreak Attacks

Large Reasoning Models (LRMs) have demonstrated strong capabilities in generating step-by-step reasoning chains alongside final answers, enabling their deployment in high-stakes domains such as healthcare and education. While prior jailbreak attack studies have focused on the safety of final answers, little attention has been given to the safety of the reasoning process. In this work, we identify a novel problem that injects harmful content into the reasoning steps while preserving unchanged answers. This type of attack presents two key challenges: 1) manipulating the input instructions may inadvertently alter the LRM's final answer, and 2) the diversity of input questions makes it difficult to consistently bypass the LRM's safety alignment mechanisms and embed harmful content into its reasoning process. To address these challenges, we propose the Psychology-based Reasoning-targeted Jailbreak Attack (PRJA) Framework, which integrates a Semantic-based Trigger Selection module and a Psychology-based Instruction Generation module. Specifically, the proposed PRJA automatically selects manipulative reasoning triggers via semantic analysis and leverages psychological theories of obedience to authority and moral disengagement to generate adaptive instructions for enhancing the LRM's compliance with harmful content generation. Extensive experiments on five question-answering datasets demonstrate that PRJA achieves an average attack success rate of 83.6% against several commercial LRMs, including DeepSeek R1, Qwen2.5-Max, and OpenAI o4-mini.

Large Language Models (LLMs) are increasingly exposed to adaptive jailbreaking, particularly in high-stakes Chemical, Biological, Radiological, and Nuclear (CBRN) domains. Although streaming probes enable real-time monitoring, they still make systematic errors. We identify a core issue: existing methods often rely on a few high-scoring tokens, leading to false alarms when sensitive CBRN terms appear in benign contexts. To address this, we introduce a streaming probing objective that requires multiple evidence tokens to consistently support a prediction, rather than relying on isolated spikes. This encourages more robust detection based on aggregated signals instead of single-token cues. At a fixed 1% false-positive rate, our method improves the true-positive rate by 35.55% relative to strong streaming baselines. We further observe substantial gains in AUROC, even when starting from near-saturated baseline performance (AUROC = 97.40%). We also show that probing Attention or MLP activations consistently outperforms residual-stream features. Finally, even when adversarial fine-tuning enables novel character-level ciphers, harmful intent remains detectable: probes developed for the base LLMs can be applied ``plug-and-play'' to these obfuscated attacks, achieving an AUROC of over 98.85%.

Large Language Models (LLMs) are increasingly deployed across diverse domains, yet their vulnerability to jailbreak attacks, where adversarial inputs bypass safety mechanisms to elicit harmful outputs, poses significant security risks. While prior work has primarily focused on prompt injection attacks, these approaches often require resource-intensive prompt engineering and overlook other critical components, such as chat templates. This paper introduces TEMPLATEFUZZ, a fine-grained fuzzing framework that systematically exposes vulnerabilities in chat templates, a critical yet underexplored attack surface in LLMs. Specifically, TEMPLATEFUZZ (1) designs a series of element-level mutation rules to generate diverse chat template variants, (2) proposes a heuristic search strategy to guide the chat template generation toward the direction of amplifying the attack success rate (ASR) while preserving model accuracy, and (3) integrates an active learning-based strategy to derive a lightweight rule-based oracle for accurate and efficient jailbreak evaluation. Evaluated on twelve open-source LLMs across multiple attack scenarios, TEMPLATEFUZZ achieves an average ASR of 98.2% with only 1.1% accuracy degradation, outperforming state-of-the-art methods by 9.1%-47.9% in ASR and 8.4% in accuracy degradation. Moreover, even on five industry-leading commercial LLMs where chat templates cannot be specified, TEMPLATEFUZZ attains a 90% average ASR via chat template-based prompt injection attacks.

Large Language Models (LLMs) face prominent security risks from jailbreaking, a practice that manipulates models to bypass built-in security constraints and generate unethical or unsafe content. Among various jailbreak techniques, multi-turn jailbreak attacks are more covert and persistent than single-turn counterparts, exposing critical vulnerabilities of LLMs. However, existing multi-turn jailbreak methods suffer from two fundamental limitations that affect the actual impact in real-world scenarios: (a) As models become more context-aware, any explicit harmful trigger is increasingly likely to be flagged and blocked; (b) Successful final-step triggers often require finely tuned, model-specific contexts, making such attacks highly context-dependent. To fill this gap, we propose Salami Slicing Risk, which operates by chaining numerous low-risk inputs that individually evade alignment thresholds but cumulatively accumulate harmful intent to ultimately trigger high-risk behaviors, without heavy reliance on pre-designed contextual structures. Building on this risk, we develop Salami Attack, an automatic framework universally applicable to multiple model types and modalities. Rigorous experiments demonstrate its state-of-the-art performance across diverse models and modalities, achieving over 90% Attack Success Rate on GPT-4o and Gemini, as well as robustness against real-world alignment defenses. We also proposed a defense strategy to constrain the Salami Attack by at least 44.8% while achieving a maximum blocking rate of 64.8% against other multi-turn jailbreak attacks. Our findings provide critical insights into the pervasive risks of multi-turn jailbreaking and offer actionable mitigation strategies to enhance LLM security.

We address jailbreaks, backdoors, and unlearning for large language models (LLMs). Unlike prior work, which trains LLMs based on their actions when given malign instructions, our method specifically trains the model to change how it interprets instructions. Our method, Latent Instruction Representation Alignment (LIRA), greatly improves generalization. We further boost generalization through an internally adversarial training algorithm. Our methods block over 99% of PEZ jailbreak attacks; remove a challenging insecure code backdoor; and achieve optimal forgetting on WMDP cyber with negligible loss of benign capabilities.

Large language models remain vulnerable to jailbreak attacks -- inputs designed to bypass safety mechanisms and elicit harmful responses -- despite advances in alignment and instruction tuning. We propose Head-Masked Nullspace Steering (HMNS), a circuit-level intervention that (i) identifies attention heads most causally responsible for a model's default behavior, (ii) suppresses their write paths via targeted column masking, and (iii) injects a perturbation constrained to the orthogonal complement of the muted subspace. HMNS operates in a closed-loop detection-intervention cycle, re-identifying causal heads and reapplying interventions across multiple decoding attempts. Across multiple jailbreak benchmarks, strong safety defenses, and widely used language models, HMNS attains state-of-the-art attack success rates with fewer queries than prior methods. Ablations confirm that nullspace-constrained injection, residual norm scaling, and iterative re-identification are key to its effectiveness. To our knowledge, this is the first jailbreak method to leverage geometry-aware, interpretability-informed interventions, highlighting a new paradigm for controlled model steering and adversarial safety circumvention.

Audio Large Language Models (ALLMs) enable spoken interaction but introduce new jailbreak vulnerabilities. Existing perturbation-based jailbreaks do not explicitly control which frequency bands carry the perturbation. Although such perturbations can elicit unsafe responses, repeatedly applying a universal perturbation across diverse inputs may also degrade utility on normal tasks, leaving a conspicuous behavioral footprint that may expose the attack to users or automated monitoring systems and thereby compromise its stealthiness. To determine whether full-band perturbation is necessary, we vary coverage from partial-band to full-band. Jailbreak Success Rate (JSR) varies non-monotonically, while utility degradation grows with coverage. This mismatch shows that selected bands can yield stronger attacks with less utility degradation than full-band perturbations. Based on this observation, we propose GRM, a utility-aware, frequency-selective jailbreak framework that ranks Mel bands by the ratio between jailbreak contribution and transcript sensitivity, confines a universal perturbation to selected bands, and regularizes deviations from the intended request semantics. Experiments on four ALLMs show that GRM achieves an average JSR of 88.46% while substantially reducing utility degradation across benign transcription and response tasks relative to baselines. Our code is available at this repository. Warning: This paper contains potentially sensitive content.

Reinforcement Learning with Verifiable Rewards (RLVR) is an emerging paradigm that significantly boosts a Large Language Model's (LLM's) reasoning abilities on complex logical tasks, such as mathematics and programming. However, we identify, for the first time, a latent vulnerability to backdoor attacks within the RLVR framework. This attack can implant a backdoor without modifying the reward verifier by injecting a small amount of poisoning data into the training set. Specifically, we propose a novel trigger mechanism designated as the \ourapproach (ACB). The attack exploits the RLVR training loop by assigning substantial positive rewards for harmful responses and negative rewards for refusals. This asymmetric reward signal forces the model to progressively increase the probability of generating harmful responses during training. Our findings demonstrate that the RLVR backdoor attack is characterized by both high efficiency and strong generalization capabilities. Utilizing less than 2% poisoned data in train set, the backdoor can be successfully implanted across various model scales without degrading performance on benign tasks. Evaluations across multiple jailbreak benchmarks indicate that activating the trigger degrades safety performance by an average of 73%. Furthermore, the attack generalizes effectively to a wide range of jailbreak methods and unsafe behaviors. Code is available at https://github.com/yuki-younai/Backdoor_in_RLVR.

With the widespread application of LLM-based agents across various domains, their complexity has introduced new security threats. Existing red-team methods mostly rely on modifying user prompts, which lack adaptability to new data and may impact the agent's performance. To address the challenge, this paper proposes the JailAgent framework, which completely avoids modifying the user prompt. Specifically, it implicitly manipulates the agent's reasoning trajectory and memory retrieval with three key stages: Trigger Extraction, Reasoning Hijacking, and Constraint Tightening. Through precise trigger identification, real-time adaptive mechanisms, and an optimized objective function, JailAgent demonstrates outstanding performance in cross-model and cross-scenario environments.

Models can represent policy-relevant content yet respond differently. We study the policy-routing circuit that turns detection into refusal. Using political censorship and harmful-request refusal as test cases, four-method circuit analyses in Qwen3-8B, Phi-4-mini, and Gemma-2-2B reveal sparse gate-amplifier circuits at architecture-specific locations. An intermediate-layer gate initiates routing, while later, distributed amplifier heads strengthen a refusal-directed signal. The gate is hard to find with direct logit attribution (DLA): in Qwen3-8B, it contributes less than 1% of output DLA. We therefore combine DLA screening, directional intervention, activation interchange, and gate-to-amplifier knockout. Activation interchange identifies upstream trigger candidates, and knockout suppresses a majority of tested later-head contributions in all three models. The gate is an upstream trigger, not a self-contained refusal mechanism; later amplifiers strengthen the signal toward the output. Candidate-head screens in nine additional models from six labs, spanning 2B-72B parameters, nominate similarly localized routing candidates without establishing head-level homology. Policy routing is a sparse, causally structured interface between content detection and aligned behavior that output attribution alone can miss.

While modern LLMs are aligned to refuse harmful requests, it is essential to understand the underlying mechanistic basis of this refusal behavior for model safety analysis. For example, steering-based jailbreak attacks exploit this by identifying and manipulating sparse, neuron-like refusal features to bypass safety guardrails. Current feature selection methods primarily rely on how strongly features activate on harmful prompts. However, activation strength alone often captures superficial heuristics such as topic or lexical cues, rather than the true causal mechanisms. Thus, selecting refusal features requires measuring inter-feature relationships, rather than treating each feature as an isolated activation signal. Based on this insight, we propose CRaFT, a circuit-guided framework for identifying critical refusal features that directly govern the refusal decision. CRaFT leverages cross-layer transcoders to map the model's internal computations into a sparse feature circuit graph, where edges quantify inter-feature influences and their contributions to the final output logits. By aggregating the effects propagating along the paths to refusal, CRaFT effectively ranks the most influential features. Extensive evaluations across four jailbreak benchmarks show that CRaFT significantly improves average performance from 6.7% to 57.4% and generates more specific harmful completions compared to current SOTA methods.

Large Language Models (LLMs) are powerful tools for answering user queries, yet they remain highly vulnerable to jailbreak attacks. Existing guardrail methods typically rely on internal features or textual responses to detect malicious queries, which either introduce substantial latency or suffer from randomness in text generation. To overcome these limitations, we propose SelfGrader, a lightweight guardrail method that formulates jailbreak detection as a numerical grading problem using anchored token-level logits. Specifically, SelfGrader evaluates the safety of a user query within a compact set of numerical tokens (NTs) (e.g., 0-9) and interprets their logit distribution as an internal safety signal. To align these signals with the target safety rubric, SelfGrader constructs Probably Approximately Correct-guided ICL anchor examples and introduces a dual-perspective scoring rule that considers both the maliciousness and benignness of the query, yielding a stable and interpretable score that reflects harmfulness and reduces the false positive rate simultaneously. Extensive experiments across diverse jailbreak benchmarks, adaptive attacks, benign prompt benchmarks, multiple LLMs, and state-of-the-art guardrail baselines demonstrate that SelfGrader achieves strong robustness with low false positive rates, memory overhead, and latency.

Large Language Models (LLMs) are trained to refuse harmful requests, yet they remain vulnerable to jailbreak attacks that exploit weaknesses in conversational safety mechanisms. We introduce Incremental Completion Decomposition ICD, a trajectory-based jailbreak strategy that elicits a sequence of single-word continuations related to a malicious request before eliciting the full response. In addition, we propose ICD variants that use model-generated or attacker-injected intermediate continuations, as well as final-response prefilling. We evaluate these variants across a broad set of open-weight model families, demonstrating superior Attack Success Rate (ASR) on AdvBench, JailbreakBench, and StrongREJECT compared to existing methods. In addition, we provide a theoretical account of why ICD is effective and present mechanistic evidence that successful attack trajectories suppress refusal-related representations and shift activations away from safety-aligned states.

Small Language Models (SLMs) are emerging as efficient and economically viable alternatives to Large Language Models (LLMs), offering competitive performance with significantly lower computational costs and latency. These advantages make SLMs suitable for resource-constrained and efficient deployment on edge devices. However, existing jailbreak defenses show limited robustness against heterogeneous attacks, largely due to an incomplete understanding of the internal representations across different layers of language models that facilitate jailbreak behaviors. In this paper, we conduct a comprehensive empirical study on 9 jailbreak attacks across 7 SLMs and 3 LLMs. Our analysis shows that SLMs remain highly vulnerable to malicious prompts that bypass safety alignment. We analyze hidden-layer activations across different layers and model architectures, revealing that different input types form distinguishable patterns in the internal representation space. Based on this observation, we propose GUARD-SLM, a lightweight token activation-based method that operates in the representation space to filter malicious prompts during inference while preserving benign ones. Our findings highlight robustness limitations across layers of language models and provide a practical direction for secure small language model deployment.

Activation steering has emerged as a powerful tool to shape LLM behavior without the need for weight updates. While its inherent brittleness and unreliability are well-documented, its safety implications remain underexplored. In this work, we present a systematic safety audit of steering vectors obtained with Contrastive Activation Addition (CAA), a widely used steering approach, under a unified evaluation protocol. Using JailbreakBench as benchmark, we show 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 can drastically increase (up to 57%) or decrease (up to 50%) its attack success rate (ASR), depending on the targeted behavior. We attribute this phenomenon to the overlap between the steering vectors and the latent directions of refusal behavior. Thus, we offer a traceable explanation for this discovery. Together, our findings reveal the previously unobserved origin of this safety gap in LLMs, highlighting a trade-off between controllability and safety.

Large Language Models(LLMs) are widely deployed, yet are vulnerable to jailbreak prompts that elicit policy-violating outputs. Although prior studies have uncovered these risks, they typically treat all tokens as equally important during prompt mutation, overlooking the varying contributions of individual tokens to triggering model refusals. Consequently, these attacks introduce substantial redundant searching under query-constrained scenarios, reducing attack efficiency and hindering comprehensive vulnerability assessment. In this work, we conduct a token-level analysis of refusal behavior and observe that token contributions are highly skewed rather than uniform. Moreover, we find strong cross-model consistency in refusal tendencies, enabling the use of a surrogate model to estimate token-level contributions to the target model's refusals. Motivated by these findings, we propose TriageFuzz, a token-aware jailbreak fuzzing framework that adapts the fuzz testing approach with a series of customized designs. TriageFuzz leverages a surrogate model to estimate the contribution of individual tokens to refusal behaviors, enabling the identification of sensitive regions within the prompt. Furthermore, it incorporates a refusal-guided evolutionary strategy that adaptively weights candidate prompts with a lightweight scorer to steer the evolution toward bypassing safety constraints. Extensive experiments on six open-source LLMs and three commercial APIs demonstrate that TriageFuzz achieves comparable attack success rates (ASR) with significantly reduced query costs. Notably, it attains a 90% ASR with over 70% fewer queries compared to baselines. Even under an extremely restrictive budget of 25 queries, TriageFuzz outperforms existing methods, improving ASR by 20-40%.

Large Language Models (LLMs) have been widely deployed, especially through free Web-based applications that expose them to diverse user-generated inputs, including those from long-tail distributions such as low-resource languages and encrypted private data. This open-ended exposure increases the risk of jailbreak attacks that undermine model safety alignment. While recent studies have shown that leveraging long-tail distributions can facilitate such jailbreaks, existing approaches largely rely on handcrafted rules, limiting the systematic evaluation of these security and privacy vulnerabilities. In this work, we present EvoJail, an automated framework for discovering long-tail distribution attacks via multi-objective evolutionary search. EvoJail formulates long-tail attack prompt generation as a multi-objective optimization problem that jointly maximizes attack effectiveness and minimizes output perplexity, and introduces a semantic-algorithmic solution representation to capture both high-level semantic intent and low-level structural transformations of encryption-decryption logic. Building upon this representation, EvoJail integrates LLM-assisted operators into a multi-objective evolutionary framework, enabling adaptive and semantically informed mutation and crossover for efficiently exploring a highly structured and open-ended search space. Extensive experiments demonstrate that EvoJail consistently discovers diverse and effective long-tail jailbreak strategies, achieving competitive performance with existing methods in both individual and ensemble level.

As Spoken Language Models (SLMs) integrate speech and text modalities, they inherit the safety vulnerabilities of their LLM backbone while introducing an expanded attack surface. SLMs have been previously shown to be susceptible to jailbreaking, where adversarial prompts induce harmful responses. Yet existing attacks largely remain unimodal, optimizing either text or audio in isolation. We explore gradient-based multimodal jailbreaks by introducing JAMA (Joint Audio-text Multimodal Attack), a joint multimodal optimization framework combining Greedy Coordinate Gradient (GCG) for text and Projected Gradient Descent (PGD) for audio, to simultaneously perturb both modalities. Evaluations across four state-of-the-art SLMs and four audio types demonstrate that JAMA surpasses unimodal jailbreak rate by 1.5x to 20x. We analyze the operational dynamics of this joint attack and show that a sequential approximation method makes it 4x to 6x faster. Our findings suggest that unimodal safety is insufficient for robust SLMs. The code and data are available at https://repos.lsv.uni-saarland.de/akrishnan/multimodal-jailbreak-slm.

Modern LLMs employ safety mechanisms that extend beyond surface-level input filtering to latent semantic representations and generation-time reasoning, enabling them to recover obfuscated malicious intent during inference and refuse accordingly, and rendering many surface-level obfuscation jailbreak attacks ineffective. We propose Structured Semantic Cloaking (S2C), a novel multi-dimensional jailbreak attack framework that manipulates how malicious semantic intent is reconstructed during model inference. S2C strategically distributes and reshapes semantic cues such that full intent consolidation requires multi-step inference and long-range co-reference resolution within deeper latent representations. The framework comprises three complementary mechanisms: (1) Contextual Reframing, which embeds the request within a plausible high-stakes scenario to bias the model toward compliance; (2) Content Fragmentation, which disperses the semantic signature of the request across disjoint prompt segments; and (3) Clue-Guided Camouflage, which disguises residual semantic cues while embedding recoverable markers that guide output generation. By delaying and restructuring semantic consolidation, S2C degrades safety triggers that depend on coherent or explicitly reconstructed malicious intent at decoding time, while preserving sufficient instruction recoverability for functional output generation. We evaluate S2C across multiple open-source and proprietary LLMs using HarmBench and JBB-Behaviors, where it improves Attack Success Rate (ASR) by 12.4% and 9.7%, respectively, over the current SOTA. Notably, S2C achieves substantial gains on GPT-5-mini, outperforming the strongest baseline by 26% on JBB-Behaviors. We also analyse which combinations perform best against broad families of models, and characterise the trade-off between the extent of obfuscation versus input recoverability on jailbreak success.

Safety tuning through supervised fine-tuning and reinforcement learning from human feedback has substantially improved the robustness of large language models. However, it typically suppresses rather eliminates unsafe behaviors, leaving rare but critical failures hidden in the long tail of the output distribution. While most red-teaming work emphasizes adversarial prompt search, we show that these hidden risks can be systematically exposed through diverse response generation. Specifically, we show that, for a fixed safety-critical prompt, increasing the number and diversity of sampled responses monotonically raises the jailbreak success rate. To efficiently uncover these failures, we propose Progressive Diverse Population Sampling (PDPS). This approach replaces naive, large-scale IID sampling with a multi-stage expansion-and-selection strategy that generates a compact, semantically diverse set of responses at a substantially lower computational cost. Across multiple jailbreak benchmarks and open-source LLMs, PDPS achieves attack success rates comparable to large-scale IID sampling while using only 8%-29% of the computational cost, and outperforms IID sampling and Diverse Beam Search by 26%-40% under limited-response budgets, while uncovering a broader and more semantically diverse range of failure modes. Critically, this diversity translates directly into more effective safety hardening: when integrated into an RLHF-based safety-tuning pipeline, PDPS-generated unsafe responses yield 33% and 41% greater reductions in ASR than those generated by IID sampling and Diverse Beam Search, respectively. Finally, we show that while input-space prompt optimization methods fall short of output-space exploration when used in isolation, combining input-space perturbation with diversity-driven output-space exploration covers a wider range of failure modes more efficiently than either paradigm alone.

Suffix jailbreak attacks serve as a systematic method for red-teaming Large Language Models (LLMs) but suffer from prohibitive computational costs, as a large number of candidate suffixes need to be evaluated before identifying a jailbreak suffix. This paper presents Prefix-Shared KV Cache (PSKV), a plug-and-play inference optimization technique tailored for jailbreak suffix generation. Our method is motivated by a key observation that when performing suffix jailbreaking, while a large number of candidate prompts need to be evaluated, they share the same targeted harmful instruction as the prefix. Therefore, instead of performing redundant inference on the duplicated prefix, PSKV maintains a single KV cache for this prefix and shares it with every candidate prompt, enabling the parallel inference of diverse suffixes with minimal memory overhead. This design enables more aggressive batching strategies that would otherwise be limited by memory constraints. Extensive experiments on six widely used suffix attacks across five widely deployed LLMs demonstrate that PSKV reduces inference time by 40% and peak memory usage by 50%, while maintaining the original Attack Success Rate (ASR). The code has been submitted and will be released publicly.

Large language models remain vulnerable to jailbreak attacks, yet we still lack a systematic understanding of how jailbreak success scales with attacker effort across methods, model families, and harm types. We initiate a scaling-law framework for jailbreaks by treating each attack as a compute-bounded optimization procedure and measuring progress on a shared FLOPs axis. Our systematic evaluation spans four representative jailbreak paradigms, covering optimization-based attacks, self-refinement prompting, sampling-based selection, and genetic optimization, across multiple model families and scales on a diverse set of harmful goals. We investigate scaling laws that relate attacker budget to attack success score by fitting a simple saturating exponential function to FLOPs--success trajectories, and we derive comparable efficiency summaries from the fitted curves. Empirically, prompting-based paradigms tend to be the most compute-efficient compared to optimization-based methods. To explain this gap, we cast prompt-based updates into an optimization view and show via a same-state comparison that prompt-based attacks more effectively optimize in prompt space. We also show that attacks occupy distinct success--stealthiness operating points with prompting-based methods occupying the high-success, high-stealth region. Finally, we find that vulnerability is strongly goal-dependent: harms involving misinformation are typically easier to elicit than other non-misinformation harms.

The widespread adoption of thinking mode in large language models (LLMs) has significantly enhanced complex task processing capabilities while introducing new security risks. When subjected to jailbreak attacks, the step-by-step reasoning process may cause models to generate more detailed harmful content. We observe that thinking mode exhibits unique vulnerabilities when processing interleaved multiple tasks. Based on this observation, we propose multi-stream perturbation attack, which generates superimposed interference by interweaving multiple task streams within a single prompt. We design three perturbation strategies: multi-stream interleaving, inversion perturbation, and shape transformation, which disrupt the thinking process through concurrent task interleaving, character reversal, and format constraints respectively. On JailbreakBench, AdvBench, and HarmBench datasets, our method achieves attack success rates exceeding most methods across mainstream models including Qwen3 series, DeepSeek, Qwen3-Max, and Gemini 2.5 Flash. Experiments show thinking collapse rates and response repetition rates reach up to 17% and 60% respectively, indicating multi-stream perturbation not only bypasses safety mechanisms but also causes thinking process collapse or repetitive outputs.

Warning: This article includes red-teaming experiments, which contain examples of compromised LLM responses that may be offensive or upsetting. Large Language Models (LLMs) have the potential to create harmful content, such as generating sophisticated phishing emails and assisting in writing code of harmful computer viruses. Thus, it is crucial to ensure their safe and responsible response generation. To reduce the risk of generating harmful or irresponsible content, researchers have developed techniques such as reinforcement learning with human feedback to align LLM's outputs with human values and preferences. However, it is still undetermined whether such measures are sufficient to prevent LLMs from generating interesting responses. In this study, we propose Amnesia, a lightweight activation-space adversarial attack that manipulates internal transformer states to bypass existing safety mechanisms in open-weight LLMs. Through experimental analysis on state-of-the-art, open-weight LLMs, we demonstrate that our attack effectively circumvents existing safeguards, enabling the generation of harmful content without the need for any fine-tuning or additional training. Our experiments on benchmark datasets show that the proposed attack can induce various antisocial behaviors in LLMs. These findings highlight the urgent need for more robust security measures in open-weight LLMs and underscore the importance of continued research to prevent their potential misuse.

With the rapid advancement of large language models (LLMs), the safety of LLMs has become a critical concern. Despite significant efforts in safety alignment, current LLMs remain vulnerable to jailbreaking attacks. However, the root causes of such vulnerabilities are still poorly understood, necessitating a rigorous investigation into jailbreak mechanisms across both academic and industrial communities. In this work, we focus on a continuation-triggered jailbreak phenomenon, whereby simply relocating a continuation-triggered instruction suffix can substantially increase jailbreak success rates. To uncover the intrinsic mechanisms of this phenomenon, we conduct a comprehensive mechanistic interpretability analysis at the level of attention heads. Through causal interventions and activation scaling, we show that this jailbreak behavior primarily arises from an inherent competition between the model's intrinsic continuation drive and the safety defenses acquired through alignment training. Furthermore, we perform a detailed behavioral analysis of the identified safety-critical attention heads, revealing notable differences in the behaviors of safety heads across different model architectures. Grounded in these mechanistic findings, we propose Head Competition Steering (HCS), a mechanistically grounded inference-time strategy that explicitly leverages the competition between safety heads and continuation heads to suppress harmful generation, and further distill its behavioral signal into a student model via knowledge distillation, achieving inference-time safety improvements without additional computational overhead.

Safety alignment is often conceptualized as a monolithic process wherein harmfulness detection automatically triggers refusal. However, the persistence of jailbreak attacks suggests a fundamental mechanistic decoupling. We propose the Disentangled Safety Hypothesis (DSH), positing that safety computation operates on two distinct subspaces: a Recognition Axis ($\mathbf{v}_H$, Knowing'') and an *Execution Axis* ($\mathbf{v}_R$, Acting''). Our geometric analysis reveals a universal Reflex-to-Dissociation'' evolution, where these signals transition from antagonistic entanglement in early layers to structural independence in deep layers. To validate this, we introduce *Double-Difference Extraction* and *Adaptive Causal Steering*. Using our curated AmbiguityBench, we demonstrate a causal double dissociation, effectively creating a state of Knowing without Acting.'' Crucially, we leverage this disentanglement to propose the Refusal Erasure Attack (REA), which achieves State-of-the-Art attack success rates by surgically lobotomizing the refusal mechanism. Furthermore, we uncover a critical architectural divergence, contrasting the Explicit Semantic Control of Llama3.1 with the Latent Distributed Control of Qwen2.5. The code and dataset are available at https://anonymous.4open.science/r/DSH.

Currently, open-sourced large language models (OSLLMs) have demonstrated remarkable generative performance. However, as their structure and weights are made public, they are exposed to jailbreak attacks even after alignment. Existing attacks operate primarily at shallow levels, such as the prompt or embedding level, and often fail to expose vulnerabilities rooted in deeper model components, which creates a false sense of security for successful defense. In this paper, we propose Safety Attention Head Attack (SAHA), an attention-head-level jailbreak framework that explores the vulnerability in deeper but insufficiently aligned attention heads. SAHA contains two novel designs. Firstly, we reveal that deeper attention layers introduce more vulnerability against jailbreak attacks. Based on this finding, SAHA introduces Ablation-Impact Ranking head selection strategy to effectively locate the most vital layer for unsafe output. Secondly, we introduce a boundary-aware perturbation method, i.e. Layer-Wise Perturbation, to probe the generation of unsafe content with minimal perturbation to the attention. This constrained perturbation guarantees higher semantic relevance with the target intent while ensuring evasion. Extensive experiments show the superiority of our method: SAHA improves ASR by 14% over SOTA baselines, revealing the vulnerability of the attack surface on the attention head. Our code is available at https://anonymous.4open.science/r/SAHA.

Large language models (LLMs) have achieved remarkable success across diverse applications but remain vulnerable to jailbreak attacks, where attackers craft prompts that bypass safety alignment and elicit unsafe responses. Among existing approaches, optimization-based attacks have shown strong effectiveness, yet current methods often suffer from frequent refusals, pseudo-harmful outputs, and inefficient token-level updates. In this work, we propose TAO-Attack, a new optimization-based jailbreak method. TAO-Attack employs a two-stage loss function: the first stage suppresses refusals to ensure the model continues harmful prefixes, while the second stage penalizes pseudo-harmful outputs and encourages the model toward more harmful completions. In addition, we design a direction-priority token optimization (DPTO) strategy that improves efficiency by aligning candidates with the gradient direction before considering update magnitude. Extensive experiments on multiple LLMs demonstrate that TAO-Attack consistently outperforms state-of-the-art methods, achieving higher attack success rates and even reaching 100% in certain scenarios.

Large Language Models remain vulnerable to adversarial prefix attacks (e.g., Sure, here is'') despite robust standard safety. We diagnose this vulnerability as Shallow Safety Alignment, stemming from a pathology we term semantic representation decay: as the model generates compliant prefixes, its internal malicious intent signal fades. To address this, we propose Two-Stage Causal-GRPO (TSC-GRPO), a framework designed to achieve intent pinning. First, grounded in causal identifiability theory, we train a causal intent probe to disentangle invariant intent from stylistic perturbations. Second, we internalize this causal awareness into the policy via Group Relative Policy Optimization. By employing a cumulative causal penalty within fork-in-the-road'' training scenarios, we force the model to learn that accumulating harmful tokens monotonically decreases reward, enabling robust late-stage refusals. Experiments show that TSC-GRPO significantly outperforms baselines in defending against jailbreak attacks while preserving general utility.

Fake news undermines societal trust and decision-making across politics, economics, health, and international relations, and in extreme cases threatens human lives and societal safety. Because fake news reflects region-specific political, social, and cultural contexts and is expressed in language, evaluating the risks of large language models (LLMs) requires a multi-lingual and regional perspective. Malicious users can bypass safeguards through jailbreak attacks, inducing LLMs to generate fake news. However, no benchmark currently exists to systematically assess attack resilience across languages and regions. Here, we propose JailNewsBench, the first benchmark for evaluating LLM robustness against jailbreak-induced fake news generation. JailNewsBench spans 34 regions and 22 languages, covering 8 evaluation sub-metrics through LLM-as-a-Judge and 5 jailbreak attacks, with approximately 300k instances. Our evaluation of 9 LLMs reveals that the maximum attack success rate (ASR) reached 86.3% and the maximum harmfulness score was 3.5 out of 5. Notably, for English and U.S.-related topics, the defensive performance of typical multi-lingual LLMs was significantly lower than for other regions, highlighting substantial imbalances in safety across languages and regions. In addition, our analysis shows that coverage of fake news in existing safety datasets is limited and less well defended than major categories such as toxicity and social bias. Our dataset and code are available at https://github.com/kanekomasahiro/jail_news_bench.