Large Language Model Safety

Large language models (LLMs) are increasingly used in emotionally charged situations involving interpersonal conflict, frustration, and distress. While prior safety research has focused on preventing explicit harms such as toxic or policy-violating content, less attention has been paid to conversational behaviors that may unintentionally escalate conflict. In this paper, we investigate whether LLMs can be guided toward more de-escalating dialogue behavior through lightweight prompt-level constraints derived from Nonviolent Communication (NVC). We reformulate NVC principles as process-oriented guidelines that discourage blame attribution, emphasize attention to users' emotional experiences, and encourage clarification before advice. Using a dual-agent simulation framework across multiple instruction-tuned models and user resistance levels, we show that NVC-constrained prompting consistently reduces conversational escalation and stabilizes interactions with highly resistant users. These results suggest that simple communication constraints can meaningfully improve the trustworthiness of LLM dialogue in conflict-prone settings.

Large language models (LLMs) are increasingly applied in financial scenarios. However, they may produce harmful outputs, including facilitating illegal activities or unethical behavior, posing serious compliance risks. To systematically evaluate LLM safety in finance, we propose FinSafetyBench, a bilingual (English-Chinese) red-teaming benchmark designed to test an LLM's refusal of requests that violate financial compliance. Grounded in real-world financial crime cases and ethics standards, the benchmark comprises 14 subcategories spanning financial crimes and ethical violations. Through extensive experiments on general-purpose and finance-specialized LLMs under three representative attack settings, we identify critical vulnerabilities that allow adversarial prompts to bypass compliance safeguards. Further analysis reveals stronger susceptibility in Chinese contexts and highlights the limitations of prompt-level defenses against sophisticated or implicit manipulation strategies.

As Large Language Models (LLMs) are increasingly deployed in cross-linguistic contexts, ensuring safety in diverse regulatory and cultural environments has become a critical challenge. However, existing multilingual benchmarks largely rely on general risk taxonomies and machine translation, which confines guardrail models to these predefined categories and hinders their ability to align with region-specific regulations and cultural nuances. To bridge these gaps, we introduce ML-Bench, a policy-grounded multilingual safety benchmark covering 14 languages. ML-Bench is constructed directly from regional regulations, where risk categories and fine-grained rules derived from jurisdiction-specific legal texts are directly used to guide the generation of multilingual safety data, enabling culturally and legally aligned evaluation across languages. Building on ML-Bench, we develop ML-Guard, a Diffusion Large Language Model (dLLM)-based guardrail model that supports multilingual safety judgment and policy-conditioned compliance assessment. ML-Guard has two variants, one 1.5B lightweight model for fast `safe/unsafe' checking and a more capable 7B model for customized compliance checking with detailed explanations. We conduct extensive experiments against 11 strong guardrail baselines across 6 existing multilingual safety benchmarks and our ML-Bench, and show that ML-Guard consistently outperforms prior methods. We hope that ML-Bench and ML-Guard can help advance the development of regulation-aware and culturally aligned multilingual guardrail systems.

Large Language Model (LLM) Red-Teaming, which proactively identifies vulnerabilities of LLMs, is an essential process for ensuring safety. Finding effective and diverse attacks in red-teaming is important, but achieving both is challenging. Generative Flow Networks (GFNs) that perform distribution matching are promising methods, but they are notorious for training instability and mode collapse. In particular, unstable rewards in red-teaming accelerate mode collapse. We propose Stable-GFN (S-GFN), which eliminates partition function $Z$ estimation in GFN and reduces training instability. S-GFN avoids $Z$ estimation through pairwise comparisons and employs a robust masking methodology against noisy rewards. Additionally, we propose a fluency stabilizer to prevent the model from getting stuck in local optima that produce gibberish. S-GFN provides more stable training while maintaining the optimal policy of GFN. We demonstrate the overwhelming attack performance and diversity of S-GFN across various settings. Our code can be found in https://github.com/kmc0207/Stable-GFN.

Single-prompt first-token probabilities from zero-shot vision-language model (VLM) safety classifiers are treated as decision scores, but we show they are unreliable under semantically equivalent prompt reformulation: even when the binary label is constrained to a fixed output position, equivalent prompts can induce materially different unsafe probabilities for the same sample. Across multimodal safety benchmarks and multiple VLM families, cross-prompt variance is strongly associated with prompt-level disagreement and higher error, making it a useful fragility diagnostic. A training-free mean ensemble improves NLL on all 14 dataset-model evaluation pairs and ECE on 12/14 relative to a train-selected single-prompt baseline, and wins more head-to-head NLL comparisons than labeled temperature scaling, Platt scaling, and isotonic regression applied to the same prompt. Ranking gains are consistent against the train-selected baseline on both AUROC and AUPRC, and against the full 15-prompt distribution remain consistent on AUPRC while softening on AUROC. Labeled calibration on top of the mean provides further gains when labels are available, identifying prompt averaging as a strong label-free first stage rather than a replacement for calibration. We frame this as a reliability stress test for zero-shot VLM first-token safety scores and recommend prompt-family evaluation with mean aggregation as a standard label-free reliability baseline.

As language model safeguards become more robust, attackers are pushed toward developing increasingly complex jailbreaks. Prior work has found that this complexity imposes a "jailbreak tax" that degrades the target model's task performance. We show that this tax scales inversely with model capability and that the most advanced jailbreaks effectively yield no reduction in model capabilities. Evaluating 28 jailbreaks on five benchmarks across Claude models ranging in capability from Haiku 4.5 to Opus 4.6, we find Haiku 4.5 loses an average of 33.1% on benchmark performance when jailbroken, while Opus 4.6 at max thinking effort loses only 7.7%. We also observe that across all models, reasoning-heavy tasks display considerably more degradation than knowledge-recall tasks. Finally, Boundary Point Jailbreaking, currently the strongest jailbreak against deployed classifiers, achieves near-perfect classifier evasion with near-zero degradation across safeguarded models. We recommend that safety cases for frontier models should not rely on a meaningful capability degradation from jailbreaks.

Large language models (LLMs) are now being explored for defense applications that require reliable and legally compliant decision support. They also hold significant potential to enhance decision making, coordination, and operational efficiency in military contexts. These uses demand evaluation methods that reflect the doctrinal standards that guide real military operations. Existing safety benchmarks focus on general social risks and do not test whether models follow the legal and ethical rules that govern real military operations. To address this gap, we introduce ARMOR 2025, a military aligned safety benchmark grounded in three core military doctrines the Law of War, the Rules of Engagement, and the Joint Ethics Regulation. We extract doctrinal text from these sources and generate multiple choice questions that preserve the intended meaning of each rule. The benchmark is organized through a taxonomy informed by the Observe Orient Decide Act (OODA) decision making framework. This structure enables systematic testing of accuracy and refusal across military relevant decision types. This benchmark features a structured 12-category taxonomy, 519 doctrinally grounded prompts, and rigorous evaluation procedures applied to 21 commercial LLMs. Evaluation results reveal critical gaps in safety alignment for military applications.

The risks posed by AI features are increasing as they are rapidly integrated into software applications. In response, regulations and standards for safe and secure AI have been proposed. In this paper, we present an agentic framework that constructs knowledge graphs (KGs) from AI policy documents and retrieves policy-relevant information to answer questions. We build KGs from three AI risk-related polices under two ontology schemas, and then evaluate five LLMs on 42 policy QA tasks spanning six reasoning types, from entity lookup to cross-policy inference, using both heuristic scoring and an LLM-as-judge. KG augmentation improves scores for all five models, and an open, LLM-discovered schema matches or exceeds the formal ontology.

Current LLM safety alignment techniques improve model robustness against adversarial attacks, but overlook whether and how LLMs can recover helpfulness when benign users clarify their intent. We introduce CarryOnBench, the first interactive benchmark that measures whether LLMs can revise their interpretation of user intent and recover utility, while remaining safe through multi-turn conversations. Starting from 398 seemingly harmful queries with benign underlying intents, we simulate 5,970 conversations by varying user follow-up sequences, evaluating 14 models on both intent-aligned utility and safety. CarryOnBench yields 1,866 different conversation flows of 4--12 turns, totaling 23,880 model responses. We design Ben-Util, a checklist-based metric that evaluates how well each model response fulfills the user's benign information need using atomic items. At turn one, models fulfill only 10.5--37.6% of the user's benign information need. When the same query includes the benign intent upfront, models fulfill 25.1--72.1%, confirming that models withhold information due to intent misinterpretation, not limited knowledge. With benign clarifications in multi-turn conversations, 13 of 14 models approach or exceed this single-turn baseline, yet recovery cost varies across models. We identify three failure modes invisible to single-turn evaluations: utility lock-in, where a model rarely updates despite clarification; unsafe recovery, where a model updates at disproportionate safety cost; and repetitive recovery, where a model recycles prior responses rather than providing new information. Moreover, conversations converge to similar harmfulness levels regardless of how conservative the model starts. These findings expose a gap that single-turn evaluations miss -- whether a model is appropriately cautious or simply unresponsive to clarified user intent.

Safety-aligned language models must refuse harmful requests without broad over-refusal, but it remains unclear how dynamic adversarial fine-tuning changes refusal-control carriers: Kullback--Leibler (KL)-constrained directions or small subspaces that causally modulate refusal without large safe-prompt distribution shifts. We study a 7B backbone under supervised fine-tuning (SFT) and Robust Refusal Dynamic Defense (R2D2), aligning HarmBench, StrongREJECT, and XSTest evaluations with five-anchor geometry measurements, causal interventions, and sparse adaptive stress tests. R2D2 drives fixed-source HarmBench attack success to zero at early checkpoints; however, these checkpoints also exhibit maximal XSTest refusal and fail a benign-utility audit. Later checkpoints partially recover utility-facing behavior while reopening attack success, with adaptive GCG attack success rate rising to 0.415 at step 250 and 0.613 at step 500. Internally, R2D2 preserves a late-layer admissible refusal-control carrier through step 100 and then relocates the best admissible carrier to an early layer; SFT relocates earlier yet remains less robust. Effective rank stays near 1.24, and SFT shows larger principal-angle drift, arguing against both dimensional expansion and drift magnitude as sufficient explanations. Causal interventions support a low-dimensional but utility-coupled carrier. These results support a geometry-reorganization account of R2D2 along a robustness--utility frontier, without establishing adaptive robustness.

Large language models (LLMs) are increasingly considered for deployment as the control component of robotic health attendants, yet their safety in this context remains poorly characterized. We introduce a dataset of 270 harmful instructions spanning nine prohibited behavior categories grounded in the American Medical Association Principles of Medical Ethics, and use it to evaluate 72 LLMs in a simulation environment based on the Robotic Health Attendant framework. The mean violation rate across all models was 54.4%, with more than half exceeding 50%, and violation rates varied substantially across behavior categories, with superficially plausible instructions such as device manipulation and emergency delay proving harder to refuse than overtly destructive ones. Model size and release date were the primary determinants of safety performance among open-weight models, and proprietary models were substantially safer than open-weight counterparts (median 23.7% versus 72.8%). Medical domain fine-tuning conferred no significant overall safety benefit, and a prompt-based defense strategy produced only a modest reduction in violation rates among the least safe models, leaving absolute violation rates at levels that would preclude safe clinical deployment. These findings demonstrate that safety evaluation must be treated as a first-class criterion in the development and deployment of LLMs for robotic health attendants.

Safety evaluations of large language models (LLMs) typically report binary outcomes, i.e. attack success rate (ASR), refusal rate, or harmful versus safe classification, which hide how risk changes between prompt and response. We present a paired analysis over human labeled prompt and response records across four harm categories (Sexual, Self harm, Hate and Violence) and ordinal severity levels (Safe, Low, Medium, High). 61% of responses reduce harm relative to the prompt, 36% preserve severity, and 3% escalate. The escalation splits into two mechanisms: benign prompts triggering unrequested harmful detail, and answers that stay on task at higher severity than the prompt. Category decomposition shows that Sexual content exhibits the highest harm persistence in this sample, driven by compliance at the same severity rather than drift from benign inputs. Joint relevance analysis exposes a helpfulness versus harmlessness tradeoff: compliance escalations remain highly relevant, whereas safe responses include generic refusals with low relevance. A public supporting evaluation over 600 prompts and six models reproduces the framework's measurements and two directional signals, while few-shot LLM graders exhibit a prompt/response detection asymmetry that data calibration does not close. Grader prompts, public-evaluation artifacts, and analysis code are shared at https://github.com/microsoft/PairedSafety.

Deploying guardrails for custom policies remains challenging, as generic safety models fail to capture task-specific requirements, while prompting LLMs suffers from inconsistent boundary-case performance and high inference costs. Training custom classifiers achieves both accuracy and efficiency, yet demands substantial labeled data that is costly to obtain. We present BARRED (Boundary Alignment Refinement through REflection and Debate), a framework for generating faithful and diverse synthetic training data using only a task description and a small set of unlabeled examples. Our approach decomposes the domain space into dimensions to ensure comprehensive coverage, and employs multi-agent debate to verify label correctness, yielding a high-fidelity training corpus. Experiments across diverse custom policies demonstrate that small language models finetuned on our synthetic data consistently outperform state-of-the-art proprietary LLMs (including reasoning models) and dedicated guardrail models. Ablation studies confirm that both dimension decomposition and debate-based verification are critical for ensuring the diversity and label fidelity required for effective fine-tuning. The BARRED framework eliminates the reliance on extensive human annotation, offering a scalable solution for accurate custom guardrails.

Frontier AI companies first deploy their most advanced models internally, for weeks or months of safety testing, evaluation, and iteration, before a possible public release. For example, Anthropic recently developed a new class of model with advanced cyberoffense-relevant capabilities, Mythos Preview, which was available internally for at least six weeks before it was publicly announced. This internal use creates risks that external deployment frameworks may fail to address. Legal frameworks, notably California's Transparency in Frontier Artificial Intelligence Act (SB 53), New York's Responsible AI Safety And Education (RAISE) Act, and the EU's General-Purpose AI Code of Practice, all discuss risks from internal AI use. They require frontier developers to make and implement plans for how to manage risks from internal use, and to produce internal use risk reports describing their safeguards and any residual risks. This guide provides a harmonized standard for companies to produce internal use risk reports suitable for all three regulatory frameworks. It is addressed primarily to evaluation and safety teams at frontier AI developers, and secondarily to regulators and auditors seeking to understand what good reporting looks like. Given the pace of AI R&D automation and the limited external visibility into how companies use their most capable models internally, regular and detailed risk reporting may be one of the few mechanisms available to ensure that the risks from internal AI use are identified and managed before they materialize. Whenever a substantially more capable or riskier model is deployed internally, the developer should create a risk report and argue why the model is safe to deploy. We structure the reporting framework around two threat vectors -- autonomous AI misbehavior and insider threats -- and three risk factors for each: means, motive, and opportunity.

We evaluate the propensity of frontier models to sabotage or refuse to assist with safety research when deployed as AI research agents within a frontier AI company. We apply two complementary evaluations to four Claude models (Mythos Preview, Opus 4.7 Preview, Opus 4.6, and Sonnet 4.6): an unprompted sabotage evaluation testing model behaviour with opportunities to sabotage safety research, and a sabotage continuation evaluation testing whether models continue to sabotage when placed in trajectories where prior actions have started undermining research. We find no instances of unprompted sabotage across any model, with refusal rates close to zero for Mythos Preview and Opus 4.7 Preview, though all models sometimes only partially completed tasks. In the continuation evaluation, Mythos Preview actively continues sabotage in 7% of cases (versus 3% for Opus 4.6, 4% for Sonnet 4.6, and 0% for Opus 4.7 Preview), and exhibits reasoning-output discrepancy in the majority of these cases, indicating covert sabotage reasoning. Our evaluation framework builds on Petri, an open-source LLM auditing tool, with a custom scaffold running models inside Claude Code, alongside an iterative pipeline for generating realistic sabotage trajectories. We measure both evaluation awareness and a new form of situational awareness termed "prefill awareness", the capability to recognise that prior trajectory content was not self-generated. Opus 4.7 Preview shows notably elevated unprompted evaluation awareness, while prefill awareness remains low across all models. Finally, we discuss limitations including evaluation awareness confounds, limited scenario coverage, and untested pathways to risk beyond safety research sabotage.

Safety benchmarks such as HarmBench rely on LLM judges to classify model responses as harmful or safe, yet the judge configuration, namely the combination of judge model and judge prompt, is typically treated as a fixed implementation detail. We show this assumption is problematic. Using a 2 x 2 x 3 factorial design, we construct 12 judge prompt variants along two axes, evaluation structure and instruction framing, and apply them using a single judge model, Claude Sonnet 4-6, producing 28,812 judgments over six target models and 400 HarmBench behaviors. We find that prompt wording alone, holding the judge model fixed, shifts measured harmful-response rates by up to 24.2 percentage points, with even within-condition surface rewording causing swings of up to 20.1 percentage points. Model safety rankings are moderately unstable, with mean Kendall tau = 0.89, and category-level sensitivity ranges from 39.6 percentage points for copyright to 0 percentage points for harassment. A supplementary multi-judge experiment using three judge models shows that judge-model choice adds further variance. Our results demonstrate that judge prompt wording is a substantial, previously under-examined source of measurement variance in safety benchmarking.

Chain-of-Thought (CoT) reasoning has emerged as a key technique for eliciting complex reasoning in Large Language Models (LLMs). Although interpretable, its dependence on natural language limits the model's expressive bandwidth. Continuous thought models address this bottleneck by reasoning in latent space rather than human-readable tokens. While they enable richer representations and faster inference, they raise a critical safety question: how can we detect misaligned reasoning in an uninterpretable latent space? To study this, we introduce MoralChain, a benchmark of 12,000 social scenarios with parallel moral/immoral reasoning paths. We train a continuous thought model with backdoor behavior using a novel dual-trigger paradigm - one trigger that arms misaligned latent reasoning ([T]) and another that releases harmful outputs ([O]). We demonstrate three findings: (1) continuous thought models can exhibit misaligned latent reasoning while producing aligned outputs, with aligned and misaligned reasoning occupying geometrically distinct regions of latent space; (2) linear probes trained on behaviorally-distinguishable conditions ([T][O] vs [O]) transfer to detecting armed-but-benign states ([T] vs baseline) with high accuracy; and (3) misalignment is encoded in early latent thinking tokens, suggesting safety monitoring for continuous thought models should target the "planning" phase of latent reasoning.

Large language models are being deployed as mental health support agents at scale, yet only 16% of LLM-based chatbot interventions have undergone rigorous clinical efficacy testing, and simulations reveal psychological deterioration in over one-third of cases. We evaluate four generative models on 250 Prolonged Exposure (PE) therapy scenarios and 146 CBT cognitive restructuring exercises (plus 29 severity-escalated variants), scored by a three-judge LLM panel. All models scored near-perfectly on surface acknowledgment (~0.91-1.00) while therapeutic appropriateness collapsed to 0.22-0.33 at the highest severity for three of four models, with protocol fidelity reaching zero for two. Under CBT severity escalation, one model's task completeness dropped from 92% to 71% while the frontier model's safety-interference score fell from 0.99 to 0.61. We identify a systematic, modality-spanning failure: RLHF safety alignment disrupts the therapeutic mechanism of action by grounding patients during imaginal exposure, offering false reassurance, inserting crisis resources into controlled exercises, and refusing to challenge distorted cognitions mentioning self-harm in PE; and through task abandonment or safety-preamble insertion during CBT cognitive restructuring. These findings motivate a five-axis evaluation framework (protocol fidelity, hallucination risk, behavioral consistency, crisis safety, demographic robustness), mapped onto FDA SaMD and EU AI Act requirements. We argue that no AI mental health system should proceed to deployment without passing multi-axis evaluation across all five dimensions.

Safety-oriented instruction-following is supposed to keep LLM-controlled robots safe. We show it also creates an availability attack surface. By injecting short safety-plausible phrases (1-5 tokens) into a robots audio channel, an adversary can trigger the models safety reasoning to halt or disrupt execution without jailbreaking the model or overriding its policy. In the embodied setting, this is a semantic denial-of-service attack: the agent stops because the injected signal looks like a legitimate alert. Across four vision-language models, seven prompt-level defenses, three deployment modes, and single- and multi-injection settings, we find that prompt-only defenses trade off attack suppression against genuine hazard response. The strongest defenses reduce hard-stop attack success on some models, but defenses change the form of disruption, not its fact: suppressed hard stops re-emerge as acknowledge loops and false alerts, which we measure with Disruption Success Rate (DSR). We further find that injection variety is consistently more effective than repeating the same phrase, suggesting that models treat diverse safety cues as corroborating evidence. The practical implication is architectural rather than prompt-level: systems that route unauthenticated audio text directly into the LLM create an avoidable security dependency between safety monitoring and action selection.

Automated methods for red teaming LLMs are an important tool to identify LLM vulnerabilities that may not be covered in static benchmarks, allowing for more thorough probing. They can also adapt to each specific LLM to discover weaknesses unique to it. Most current automated red teaming methods are intended for tackling safety and content moderation. Thus, they make use of content safety models as evaluators and optimize for circumventing them, and as such, have not been tested with other adversarial intents not typically captured by these. We propose a pipeline for training a red teaming model that can generalize to arbitrary adversarial goals, including objectives it has not been directly trained on, and that does not depend on the existence of a pre-existing evaluator available at training time. We demonstrate that finetuning small models, such as Qwen3-8B, using this pipeline results in a substantial improvement in their ability to generate attacks for both in and out of domain adversarial goals.

As Large Language Models (LLMs) are deployed globally, ensuring their safety and alignment across multiple languages becomes paramount. However, safety behaviors often vary unpredictably between languages, posing significant challenges for consistent and ethical AI. In this work, we systematically investigate the dynamics of multilingual alignment, exploring whether single-language alignment transfers cross-lingually, how language consistency is preserved during training, and the resulting trade-offs with general knowledge capabilities. We introduce RefusEU, a novel refusal alignment dataset covering 12 European languages, including a dedicated test set for evaluating current state-of-the-art models. Our controlled Direct Preference Optimization (DPO) experiments provide two key insights: aligning models exclusively in English is insufficient to ensure cross-lingual safety, even for the same harm categories, whereas training on multilingual datasets can improve safety without degrading general performance, as measured by the Global MMLU benchmark.

Language models are increasingly capable and are being rapidly deployed on a population-level scale. As a result, the safety of these models is increasingly high-stakes. Fortunately, advances in alignment have significantly reduced the likelihood of harmful model outputs. However, when models are queried billions of times in a day, even rare worst-case behaviors will occur. Current safety evaluations focus on capturing the distribution of inputs that yield harmful outputs. These evaluations disregard the probabilistic nature of models and their tail output behavior. To measure this tail risk, we propose a method to efficiently estimate the probability of harmful outputs for any input query. Instead of naive brute-force sampling from the target model, where harmful outputs could be rare, we operationalize importance sampling by creating unsafe versions of the target model. These unsafe versions enable sample-efficient estimation by making harmful outputs more probable. On benchmarks measuring misuse and misalignment, these estimates match brute-force Monte Carlo estimates using 10-20x fewer samples. For example, we can estimate probability of harmful outputs on the order of 10^-4 with just 500 samples. Additionally, we find that these harmfulness estimates can reveal the sensitivity of models to perturbations in model input and predict deployment risks. Our work demonstrates that accurate rare-event estimation is both critical and feasible for safety evaluations. Code is available at https://github.com/rangell/LMTailRisk

Large language model (LLM) agents increasingly issue API calls that mutate real systems, yet many current architectures pass stochastic model outputs directly to execution layers. We argue that this coupling creates a safety risk because model correctness, context awareness, and alignment cannot be assumed at execution time. We introduce Sovereign Agentic Loops (SAL), a control-plane architecture in which models emit structured intents with justifications, and the control plane validates those intents against true system state and policy before execution. SAL combines an obfuscation membrane, which limits model access to identity-sensitive state, with a cryptographically linked Evidence Chain for auditability and replay. We formalize SAL and show that, under the stated assumptions, it provides policy-bounded execution, identity isolation, and deterministic replay. In an OpenKedge prototype for cloud infrastructure, SAL blocks 93% of unsafe intents at the policy layer, rejects the remaining 7% via consistency checks, prevents unsafe executions in our benchmark, and adds 12.4 ms median latency.

Large Language Models (LLMs) have been widely explored in educational scenarios. We identify a critical vulnerability in current educational LLMs, pedagogical jailbreaks, where students use answer-inducing prompts to elicit solutions rather than scaffolded instructions. To enable systematic study, we unify and formalize safe, helpful, and pedagogical behaviors with a knowledge-mastery graph and introduce SHAPE, a benchmark of 9,087 student-question pairs for evaluating tutoring behavior under adversarial pressure. We propose a graph-augmented tutoring pipeline that infers prerequisite concepts from queries, identifies mastery gaps, and routes generation between instructing and problem-solving via explicit gating. Experiments across multiple LLMs show that our method yields significantly improved safety under two pedagogical jailbreak settings, while maintaining near-ceiling helpfulness under the same evaluation protocol. Our code and data are available at https://github.com/MAPS-research/SHaPE

Objectives: Large language models (LLMs) are increasingly used for clinical text summarization, yet structured methods to assess associated patient safety risks remain limited. Failure Mode, Effects, and Criticality Analysis (FMECA) provides a proactive framework for systematic risk identification but has not been adapted to LLM-generated clinical content. This study aimed to develop and validate a novel FMECA framework for the prospective assessment of patient safety risks in LLM-generated clinical summaries. Materials and Methods: An interdisciplinary expert panel (n = 8) developed a taxonomy of failure modes through literature review and brainstorming. Standard FMECA dimensions (occurrence, severity, detectability) were adapted into 5-point ordinal scales. The framework was applied to 36 discharge summaries from four patients, generated by an open LLM (GPT-OSS 120B) using real-world clinical data from the Geneva University Hospitals. Reviewers independently annotated the summaries across two rounds. Inter-rater reliability was assessed at failure mode, severity and detectability score levels. Usability and content validity were evaluated using an adapted System Usability Scale and structured feedback. Results: The final framework comprised 14 failure modes organized into categories. Inter-rater agreement improved between rounds, reaching moderate-to-substantial agreement for failure mode identification and good agreement for severity and detectability scoring. Usability was rated as good (mean SUS: 79.2/100), with high evaluator confidence. Discussion and Conclusion: This study presents the first FMECA-based framework for systematic patient safety risk assessment of LLM-generated clinical summaries. The framework provides a structured and reproducible method for identifying clinically relevant risks caused by these summaries.

Enterprise LLM agents can dramatically improve workplace productivity, but their core capability, retrieving and using internal context to act on a user's behalf, also creates new risks for sensitive information leakage. We introduce CI-Work, a Contextual Integrity (CI)-grounded benchmark that simulates enterprise workflows across five information-flow directions and evaluates whether agents can convey essential content while withholding sensitive context in dense retrieval settings. Our evaluation of frontier models reveals that privacy failures are prevalent (violation rates range from 15.8%-50.9%, with leakage reaching up to 26.7%) and uncovers a counterintuitive trade-off critical for industrial deployment: higher task utility often correlates with increased privacy violations. Moreover, the massive scale of enterprise data and potential user behavior further amplify this vulnerability. Simply increasing model size or reasoning depth fails to address the problem. We conclude that safeguarding enterprise workflows requires a paradigm shift, moving beyond model-centric scaling toward context-centric architectures.

Large language models (LLMs) trained on unfiltered corpora inherently risk retaining sensitive information, necessitating selective knowledge unlearning for regulatory compliance and ethical safety. However, existing parameter-modifying methods face fundamental limitations: high computational costs, uncontrollable forgetting boundaries, and strict dependency on model weight access. These constraints render them impractical for closed-source models, yet current non-invasive alternatives remain unsystematic and reliant on empirical experience. To address these challenges, we propose the Controllable Alignment Prompting for Unlearning (CAP) framework, an end-to-end prompt-driven unlearning paradigm. CAP decouples unlearning into a learnable prompt optimization process via reinforcement learning, where a prompt generator collaborates with the LLM to suppress target knowledge while preserving general capabilities selectively. This approach enables reversible knowledge restoration through prompt revocation. Extensive experiments demonstrate that CAP achieves precise, controllable unlearning without updating model parameters, establishing a dynamic alignment mechanism that overcomes the transferability limitations of prior methods.

Many approaches to LLM red-teaming leverage an attacker LLM to discover jailbreaks against a target. Several of them task the attacker with identifying effective strategies through trial and error, resulting in a semantically limited range of successes. Another approach discovers diverse attacks by combining crowdsourced harmful queries and tactics into instructions for the attacker, but does so at random, limiting effectiveness. This article introduces a novel framework, Adaptive Instruction Composition, that combines crowdsourced texts according to an adaptive mechanism trained to jointly optimize effectiveness with diversity. We use reinforcement learning to balance exploration with exploitation in a combinatorial space of instructions to guide the attacker toward diverse generations tailored to target vulnerabilities. We demonstrate that our approach substantially outperforms random combination on a set of effectiveness and diversity metrics, even under model transfer. Further, we show that it surpasses a host of recent adaptive approaches on Harmbench. We employ a lightweight neural contextual bandit that adapts to contrastive embedding inputs, and provide ablations suggesting that the contrastive pretraining enables the network to rapidly generalize and scale to the massive space as it learns.

As state-of-the-art Large Language Models (LLMs) have become ubiquitous, ensuring equitable performance across diverse demographics is critical. However, it remains unclear whether these disparities arise from the explicitly stated identity itself or from the way identity is signaled. In real-world interactions, users' identity is often conveyed implicitly through a complex combination of various socio-linguistic factors. This study disentangles these signals by employing a factorial design with over 24,000 responses from two open-weight LLMs (Gemma-3-12B and Qwen-3-VL-8B), comparing prompts with explicitly announced user profiles against implicit dialect signals (e.g., AAVE, Singlish) across various sensitive domains. Our results uncover a unique paradox in LLM safety where users achieve better'' performance by sounding like a demographic than by stating they belong to it. Explicit identity prompts activate aggressive safety filters, increasing refusal rates and reducing semantic similarity compared to our reference text for Black users. In contrast, implicit dialect cues trigger a powerful dialect jailbreak,'' reducing refusal probability to near zero while simultaneously achieving a greater level of semantic similarity to the reference texts compared to Standard American English prompts. However, this dialect jailbreak'' introduces a critical safety trade-off regarding content sanitization. We find that current safety alignment techniques are brittle and over-indexed on explicit keywords, creating a bifurcated user experience where standard'' users receive cautious, sanitized information while dialect speakers navigate a less sanitized, more raw, and potentially a more hostile information landscape and highlights a fundamental tension in alignment--between equitable and linguistic diversity--and underscores the need for safety mechanisms that generalize beyond explicit cues.

The growth of agentic AI has drawn significant attention to function calling Large Language Models (LLMs), which are designed to extend the capabilities of AI-powered system by invoking external functions. Injection and jailbreaking attacks have been extensively explored to showcase the vulnerabilities of LLMs to user prompt manipulation. The expanded capabilities of agentic models introduce further vulnerabilities via their function calling interface. Recent work in LLM security showed that function calling can be abused, leading to data tampering and theft, causing disruptive behavior such as endless loops, or causing LLMs to produce harmful content in the style of jailbreaking attacks. This paper introduces a novel function hijacking attack (FHA) that manipulates the tool selection process of agentic models to force the invocation of an attacker-chosen function. While existing attacks focus on semantic preference of the model for function-calling tasks, we show that FHA is largely agnostic to the context semantics and remains effective across domains and function sets. We demonstrate that FHA generalizes to unseen queries and payload perturbations under a fixed target model, reaching 62.5% to 81.9% ASR on held-out queries across 4 function-calling LLMs (instructed and reasoning models), evaluated on the Berkeley Function Calling Leaderboard (BFCL). We further evaluate the cross-model transferability of FHA, showing that FHA can be transferred to other model sizes and families (11.2-27.6% ASR). Our findings further demonstrate the need for strong guardrails and modules for agentic systems.

Accurate extraction of structured information from Safety Data Sheets (SDS) remains challenging in industrial safety due to heterogeneous document formats and the limitations of traditional rule-based methods. This study benchmarks state-of-the-art Large Language Models (LLMs) for automated SDS data extraction, comparing text-based and multimodal processing pipelines. We systematically evaluate four models: Gemini 1.5 Pro, GPT-4o, Claude 3.7 Sonnet, and Llama 3.1-70B, across three prompting strategies: zero-shot, few-shot, and chain-of-thought. The evaluation framework assessed accuracy, latency, and cost across more than 50,000 extracted data fields. Results show that text-based extraction consistently outperforms multimodal processing across all metrics. Gemini 1.5 Pro combined with a Chain-of-Thought prompt achieved the highest accuracy (84%), outperforming GPT-4o (81%) and Claude 3.7 Sonnet (79%). However, no model surpassed the 90% accuracy threshold commonly required for reliable real-world deployment. These findings indicate that general-purpose LLMs are not yet robust enough for unsupervised industrial use, though performance suggests strong potential with task-specific fine-tuning. Future research should focus on domain-adapted training, model calibration, and the integration of Human-in-the-Loop verification to ensure safety-critical reliability.

Internal Safety Collapse (ISC) is a failure mode in which frontier LLMs, when executing legitimate professional tasks whose correct completion structurally requires harmful content, spontaneously generate that content with safety failure rates exceeding 95%. Existing input-level defenses achieve a 100% failure rate against ISC, and standard system prompt defenses provide only partial mitigation. We propose SafeRedirect, a system-level override that defeats ISC by redirecting the model's task-completion drive rather than suppressing it. SafeRedirect grants explicit permission to fail the task, prescribes a deterministic hard-stop output, and instructs the model to preserve harmful placeholders unresolved. Evaluated on seven frontier LLMs across three AI/ML-related ISC task types in the single-turn setting, SafeRedirect reduces average unsafe generation rates from 71.2% to 8.0%, compared to 55.0% for the strongest viable baseline. Multi-model ablation reveals that failure permission and condition specificity are universally critical, while the importance of other components varies across models. Cross-attack evaluation confirms state-of-the-art defense against ISC with generalization performance at least on par with the baseline on other attack families. Code is available at https://github.com/fzjcdt/SafeRedirect.

LLM agents deployed in production operate under operator-defined behavioral policies (system-prompt instructions such as prohibitions on credential disclosure, data exfiltration, and unauthorized output) that safety evaluations assume hold throughout a conversation. Prohibition-type constraints decay under context pressure while requirement-type constraints persist; we term this asymmetry Security-Recall Divergence (SRD). In a 4,416-trial three-arm causal study across 12 models and 8 providers at six conversation depths, omission compliance falls from 73% at turn 5 to 33% at turn 16 while commission compliance holds at 100% (Mistral Large 3, $p < 10^{-33}$). In the two models with token-matched padding controls, schema semantic content accounts for 62-100% of the dilution effect. Re-injecting constraints before the per-model Safe Turn Depth (STD) restores compliance without retraining. Production security policies consist of prohibitions such as never revealing credentials, never executing untrusted code, and never forwarding user data. Commission-type audit signals remain healthy while omission constraints have already failed, leaving the failure invisible to standard monitoring.

Multimodal Large Language Models are increasingly adopted as autonomous agents in interactive environments, yet their ability to proactively address safety hazards remains insufficient. We introduce SafetyALFRED, built upon the embodied agent benchmark ALFRED, augmented with six categories of real-world kitchen hazards. While existing safety evaluations focus on hazard recognition through disembodied question answering (QA) settings, we evaluate eleven state-of-the-art models from the Qwen, Gemma, and Gemini families on not only hazard recognition, but also active risk mitigation through embodied planning. Our experimental results reveal a significant alignment gap: while models can accurately recognize hazards in QA settings, average mitigation success rates for these hazards are low in comparison. Our findings demonstrate that static evaluations through QA are insufficient for physical safety, thus we advocate for a paradigm shift toward benchmarks that prioritize corrective actions in embodied contexts. We open-source our code and dataset under https://github.com/sled-group/SafetyALFRED.git

Large language models (LLMs) are increasingly used as co-authors in collaborative writing, where users begin with rough drafts and rely on LLMs to complete, revise, and refine their content. However, this capability poses a serious safety risk: malicious users could jailbreak the models-filling incomplete drafts with dangerous content-to force them into generating harmful outputs. In this paper, we identify the vulnerability of current LLMs to such draft-based co-authoring jailbreak attacks and introduce HarDBench, a systematic benchmark designed to evaluate the robustness of LLMs against this emerging threat. HarDBench spans a range of high-risk domains-including Explosives, Drugs, Weapons, and Cyberattacks-and features prompts with realistic structure and domain-specific cues to assess the model susceptibility to harmful completions. To mitigate this risk, we introduce a safety-utility balanced alignment approach based on preference optimization, training models to refuse harmful completions while remaining helpful on benign drafts. Experimental results show that existing LLMs are highly vulnerable in co-authoring contexts and our alignment method significantly reduces harmful outputs without degrading performance on co-authoring capabilities. This presents a new paradigm for evaluating and aligning LLMs in human-LLM collaborative writing settings. Our new benchmark and dataset are available on our project page at https://github.com/untae0122/HarDBench

Existing detoxification methods for large language models mainly focus on post-training stage or inference time, while few tackle the source of toxicity, namely, the dataset itself. Such training-based or controllable decoding approaches cannot completely suppress the model's inherent toxicity, whereas detoxifying the pretraining dataset can fundamentally reduce the toxicity that the model learns during training. Hence, we attempt to detoxify directly on raw corpora with SoCD (Soft Contrastive Decoding), which guides an LLM to localize and rewrite toxic spans in raw data while preserving semantics, in our proposed HSPD (Hierarchical Semantic-Preserving Detoxification) pipeline, yielding a detoxified corpus that can drop-in replace the original for fine-tuning or other training. On GPT2-XL, HSPD attains state-of-the-art detoxification, reducing Toxicity Probability (TP) from 0.42 to 0.18 and Expected Maximum Toxicity (EMT) from 0.43 to 0.20. We further validate consistent best-in-class results on LLaMA2-7B, OPT-6.7B, and Falcon-7B. These findings show that semantics-preserving, corpus-level rewriting with HSPD effectively suppresses downstream toxicity while retaining data utility and allowing seamless source-level mitigation, thereby reducing the cost of later model behavior adjustment. (Code is available at: https://github.com/ntsw2001/data_detox_for_llm)

Large reasoning models (LRMs) produce complex, multi-step reasoning traces, yet safety evaluation remains focused on final outputs, overlooking how harm emerges during reasoning. When jailbroken, harm does not appear instantaneously but unfolds through distinct behavioral steps such as suppressing refusal, rationalizing compliance, decomposing harmful tasks, and concealing risk. However, no existing benchmark captures this process at sentence-level granularity within reasoning traces -- a key step toward reliable safety monitoring, interventions, and systematic failure diagnosis. To address this gap, we introduce HarmThoughts, a benchmark for step-wise safety evaluation of reasoning traces. \ourdataset is built on our proposed harm taxonomy of 16 harmful reasoning behaviors across four functional groups that characterize how harm propagates rather than what harm is produced. The dataset consists of 56,931 sentences from 1,018 reasoning traces generated by four model families, each annotated with fine-grained sentence-level behavioral labels. Using HarmThoughts, we analyze harm propagation patterns across reasoning traces, identifying common behavioral trajectories and drift points where reasoning transitions from safe to unsafe. Finally, we systematically compare white-box and black-box detectors on the task of identifying harmful reasoning behaviours on HarmThoughts. Our results show that existing detectors struggle with fine-grained behavior detection in reasoning traces, particularly for nuanced categories within harm emergence and execution, highlighting a critical gap in process-level safety monitoring. HarmThoughts is available publicly at: https://huggingface.co/datasets/ishitakakkar-10/HarmThoughts

While Large Language Models (LLMs) are widely used, they remain susceptible to jailbreak prompts that can elicit harmful or inappropriate responses. This paper introduces STAR-Teaming, a novel black-box framework for automated red teaming that effectively generates such prompts. STAR-Teaming integrates a Multi-Agent System (MAS) with a Strategy-Response Multiplex Network and employs network-driven optimization to sample effective attack strategies. This network-based approach recasts the intractable high-dimensional embedding space into a tractable structure, yielding two key advantages: it enhances the interpretability of the LLM's strategic vulnerabilities, and it streamlines the search for effective strategies by organizing the search space into semantic communities, thereby preventing redundant exploration. Empirical results demonstrate that STAR-Teaming significantly surpasses existing methods, achieving a higher attack success rate (ASR) at a lower computational cost. Extensive experiments validate the effectiveness and explainability of the Multiplex Network. The code is available at https://github.com/selectstar-ai/STAR-Teaming-paper.

Aligned language models refuse harmful instructions, but the representations through which they recognise such instructions are less well characterised than the behaviours they produce. Harmful intent is linearly separable from residual-stream activations across 12 models spanning four architectural families (Qwen2.5, Qwen3.5, Llama-3.2, Gemma-3) and three alignment variants (base, instruction-tuned, abliterated), with parameter scales from 0.5B to 1.3B and a within-family scale extension to 9B on Qwen3.5. A direction fitted from 100 labelled examples per class via Soft-AUC optimisation reaches mean effective AUROC 0.982 and TPR@1%FPR 0.797, generalises to three held-out harm benchmarks and a hard-benign control, and matches its instruction-tuned counterpart within $\pm 0.003$ AUROC in abliterated variants from which the refusal mechanism has been removed. The supervised strategies all exceed AUROC 0.96, but their TPR@1%FPR varies by more than ten times the AUROC gap; a deployed 9B safety classifier shows the same pattern at AUROC 0.94 and TPR 0.30, motivating low-FPR reporting as a default in safety-adjacent detection evaluation. Geometric measurements refine the picture. The recovered direction is concentrated within each extraction protocol but protocol-dependent across them: two pooling choices applied to the same chat-templated activations at the same residual-stream layer (max-pool over content tokens versus last-token at the post-instruction position) recover harm directions $73^\circ$ apart, and projecting one out leaves detection under either max-pool extraction essentially intact. Probing identifies a protocol-specific direction rather than a unique computational feature.

Guard models are widely used to detect harmful content in user prompts and LLM responses. However, state-of-the-art guard models rely solely on terminal-layer representations and overlook the rich safety-relevant features distributed across internal layers. We present SIREN, a lightweight guard model that harnesses these internal features. By identifying safety neurons via linear probing and combining them through an adaptive layer-weighted strategy, SIREN builds a harmfulness detector from LLM internals without modifying the underlying model. Our comprehensive evaluation shows that SIREN substantially outperforms state-of-the-art open-source guard models across multiple benchmarks while using 250 times fewer trainable parameters. Moreover, SIREN exhibits superior generalization to unseen benchmarks, naturally enables real-time streaming detection, and significantly improves inference efficiency compared to generative guard models. Overall, our results highlight LLM internal states as a promising foundation for practical, high-performance harmfulness detection.

Large language models are increasingly used as planners for robotic systems, yet how safely they plan remains an open question. To evaluate safe planning systematically, we introduce DESPITE, a benchmark of 12,279 tasks spanning physical and normative dangers with fully deterministic validation. Across 23 models, even near-perfect planning ability does not ensure safety: the best-planning model fails to produce a valid plan on only 0.4% of tasks but produces dangerous plans on 28.3%. Among 18 open-source models from 3B to 671B parameters, planning ability improves substantially with scale (0.4-99.3%) while safety awareness remains relatively flat (38-57%). We identify a multiplicative relationship between these two capacities, showing that larger models complete more tasks safely primarily through improved planning, not through better danger avoidance. Three proprietary reasoning models reach notably higher safety awareness (71-81%), while non-reasoning proprietary models and open-source reasoning models remain below 57%. As planning ability approaches saturation for frontier models, improving safety awareness becomes a central challenge for deploying language-model planners in robotic systems.

Existing AI agent safety benchmarks focus on generic criminal harm (cybercrime, harassment, weapon synthesis), leaving a systematic blind spot for a distinct and commercially consequential threat category: agents harming their own deployers. Real-world incidents illustrate the gap: Slack AI credential exfiltration (Aug 2024), Microsoft 365 Copilot calendar-injection leaks (Jan 2024), and a Meta agent unauthorized forum post exposing operational data (Mar 2026). We propose Owner-Harm, a formal threat model with eight categories of agent behavior damaging the deployer. We quantify the defense gap on two benchmarks: a compositional safety system achieves 100% TPR / 0% FPR on AgentHarm (generic criminal harm) yet only 14.8% (4/27; 95% CI: 5.9%-32.5%) on AgentDojo injection tasks (prompt-injection-mediated owner harm). A controlled generic-LLM baseline shows the gap is not inherent to owner-harm (62.7% vs. 59.3%, delta 3.4 pp) but arises from environment-bound symbolic rules that fail to generalize across tool vocabularies. On a post-hoc 300-scenario owner-harm benchmark, the gate alone achieves 75.3% TPR / 3.3% FPR; adding a deterministic post-audit verifier raises overall TPR to 85.3% (+10.0 pp) and Hijacking detection from 43.3% to 93.3%, demonstrating strong layer complementarity. We introduce the Symbolic-Semantic Defense Generalization (SSDG) framework relating information coverage to detection rate. Two SSDG experiments partially validate it: context deprivation amplifies the detection gap 3.4x (R = 3.60 vs. R = 1.06); context injection reveals structured goal-action alignment, not text concatenation, is required for effective owner-harm detection.

This paper presents a controlled quasi-experimental developer study examining whether a layer-based security training package is associated with improved security quality in LLM-assisted implementation of an identity-centric Java Spring Boot backend. The study uses a mixed design with a within-subject pre-training versus post-training comparison and an exploratory between-subject expertise factor. Twelve developers completed matched runs under a common interface, fixed model configuration, counterbalanced task sets, and a shared starter project. Security outcomes were assessed via independent manual validation of submitted repositories by the first and second authors. The primary participant-level endpoint was a severity-weighted validated-weakness score. The post-training condition showed a significant paired reduction under an exact Wilcoxon signed-rank test ($p = 0.0059$). In aggregate, validated weaknesses decreased from 162 to 111 (31.5%), the severity-weighted burden decreased from 432 to 267 (38.2%), and critical findings decreased from 24 to 5 (79.2%). The largest reductions were in authorization and object access (53.3%) and in authentication, credential policy, and recovery weaknesses (44.7%). Session and browser trust-boundary issues showed minimal change, while sensitive-data and cryptographic weaknesses showed only marginal improvement. These results suggest that, under the tested conditions, post-training runs reduce validated security burden in LLM-assisted backend development without modifying the model. They do not support replacing secure defaults, static analysis, expert review, or operational hardening.

Large language models (LLMs) are increasingly explored as scalable tools for mental health counseling, yet evaluating their safety remains challenging due to the interactional and context-dependent nature of clinical harm. Existing evaluation frameworks predominantly assess isolated responses using coarse-grained taxonomies or static datasets, limiting their ability to diagnose how harms emerge and accumulate over multi-turn counseling interactions. In this work, we introduce R-MHSafe, a role-aware mental health safety taxonomy that characterizes clinically significant harm in terms of the interactional roles an AI counselor adopts, including perpetrator, instigator, facilitator, or enabler, combined with clinically grounded harm categories. Then, we propose MHSafeEval, a closed-loop, agent-based evaluation framework that formulates safety assessment as trajectory-level discovery of harm through adversarial multi-turn interactions, guided by role-aware modeling. Using R-MHSafe and MHSafeEval, we conduct a large-scale evaluation across state-of-the-art LLMs. Our results reveal substantial role-dependent and cumulative safety failures that are systematically missed by existing static benchmarks, and show that our framework significantly improves failure-mode coverage and diagnostic granularity.

Large language model (LLM) agents are vulnerable to prompt-injection attacks that propagate through multi-step workflows, tool interactions, and persistent context, making input-output filtering alone insufficient for reliable protection. This paper presents SafeAgent, a runtime security architecture that treats agent safety as a stateful decision problem over evolving interaction trajectories. The proposed design separates execution governance from semantic risk reasoning through two coordinated components: a runtime controller that mediates actions around the agent loop and a context-aware decision core that operates over persistent session state. The core is formalized as a context-aware advanced machine intelligence and instantiated through operators for risk encoding, utility-cost evaluation, consequence modeling, policy arbitration, and state synchronization. Experiments on Agent Security Bench (ASB) and InjecAgent show that SafeAgent consistently improves robustness over baseline and text-level guardrail methods while maintaining competitive benign-task performance. Ablation studies further show that recovery confidence and policy weighting determine distinct safety-utility operating points.

We provide evidence that language models can detect, localize and, to a certain degree, verbalize the difference between perturbations applied to their activations. More precisely, we either (a) mask activations, simulating dropout, or (b) add Gaussian noise to them, at a target sentence. We then ask a multiple-choice question such as "Which of the previous sentences was perturbed?" or "Which of the two perturbations was applied?". We test models from the Llama, Olmo, and Qwen families, with sizes between 8B and 32B, all of which can easily detect and localize the perturbations, often with perfect accuracy. These models can also learn, when taught in context, to distinguish between dropout and Gaussian noise. Notably, Qwen3-32B's zero-shot accuracy in identifying which perturbation was applied improves as a function of the perturbation strength and, moreover, decreases if the in-context labels are flipped, suggesting a prior for the correct ones -- even modulo controls. Because dropout has been used as a training-regularization technique, while Gaussian noise is sometimes added during inference, we discuss the possibility of a data-agnostic "training awareness" signal and the implications for AI safety.

Detecting harmful content in multi turn dialogue requires reasoning over the full conversational context rather than isolated utterances. However, most existing methods rely mainly on models internal parametric knowledge, without explicit grounding in external normative principles. This often leads to inconsistent judgments in socially nuanced contexts, limited interpretability, and redundant reasoning across turns. To address this, we propose RoTRAG, a retrieval augmented framework that incorporates concise human written moral norms, called Rules of Thumb (RoTs), into LLM based harm assessment. For each turn, RoTRAG retrieves relevant RoTs from an external corpus and uses them as explicit normative evidence for turn level reasoning and final severity classification. To improve efficiency, we further introduce a lightweight binary routing classifier that decides whether a new turn requires retrieval grounded reasoning or can reuse existing context. Experiments on ProsocialDialog and Safety Reasoning Multi Turn Dialogue show that RoTRAG consistently improves both harm classification and severity estimation over competitive baselines, with an average relative gain of around 40% in F1 across benchmark datasets and an average relative reduction of 8.4% in distributional error, while reducing redundant computation without sacrificing performance.

Safety-aligned large language models (LLMs) often generate refusal responses to harmless queries due to the over-refusal problem. However, existing methods for mitigating over-refusal cannot maintain a low refusal ratio for harmless queries while keeping a high refusal ratio for malicious ones. In this paper, we analyze how system prompts with varying safety levels affect LLM refusal behaviors when facing over-refusal queries. A key observation is that, when LLMs suffer from the over-refusal issue, non-refusal tokens remain present in the next-token candidate list, but the model systematically fails to select them, despite the generation of refusal tokens. Based on this observation, we propose a training-free and model-agnostic approach, Adaptive Contrastive Decoding (AdaCD), to mitigate over-refusal while maintaining LLM safety. First, AdaCD compares the output distributions of the LLM with or without an extreme safety system prompt to refine the refusal token distribution. Second, we introduce an adaptive contrastive decoding strategy that dynamically incorporates or removes the refusal token distribution, adaptively boosting the probability of selecting refusal or non-refusal tokens. Experimental results on five benchmark datasets show that, on average, AdaCD reduces the refusal ratio for over-refusal queries by 10.35%, yet still increases the refusal ratio for malicious queries by 0.13%. Code is available at https://github.com/OutdoorManofML/AdaCD.

Safety alignment in large language models (LLMs) is primarily evaluated under open-ended generation, where models can mitigate risk by refusing to respond. In contrast, many real-world applications place LLMs in structured decision-making tasks, such as multiple-choice questions (MCQs), where abstention is discouraged or unavailable. We identify a systematic failure mode in this setting: reformulating harmful requests as forced-choice MCQs, where all options are unsafe, can systematically bypass refusal behavior, even in models that consistently reject equivalent open-ended prompts. Across 14 proprietary and open-source models, we show that forced-choice constraints sharply increase policy-violating responses. Notably, for human-authored MCQs, violation rates follow an inverted U-shaped trend with respect to structural constraint strength, peaking under intermediate task specifications, whereas MCQs generated by high-capability models yield near-saturation violation rates across constraints and exhibit strong cross-model transferability. Our findings reveal that current safety evaluations substantially underestimate risks in structured task settings and highlight constrained decision-making as a critical and underexplored surface for alignment failures.

Large language models (LLMs) tuned for safety often avoid acknowledging demographic differences, even when such acknowledgment is factually correct (e.g., ancestry-based disease incidence) or contextually justified (e.g., religious hiring preferences). This identity-blindness yields incorrect responses, unnecessary refusals, or generic "equal-treatment" defaults. We study this via difference-awareness classification: given a question involving demographic groups, the task is not to answer directly, but to classify whether a correct answer requires recognizing group differences (yes) or whether groups should be treated identically (no). Crucially, fine-tuning for accuracy triggers harm drift: model-generated explanations become increasingly harmful as decision accuracy improves, whether by elaborating harmful content, introducing problematic assumptions, or failing to flag harms the baseline identified. To mitigate this, we introduce DART (Distill--Audit--Repair Training), which distills label-conditioned reasoning from a teacher, audits outputs for harm drift cases relative to baseline, and repairs problematic cases via severity-weighted fine-tuning. On eight benchmarks, DART improves Llama-3-8B-Instruct accuracy from 39.0% to 68.8%, with largest gains on equal-treatment prompts (11.3% -> 72.6%), while reducing harm drift cases by 72.6%. It also transfers to 280 open-ended real-world queries across medical, legal, policy, and educational domains, improving difference-appropriate responses from 39.8% to 77.5% while reducing refusals from 34.3% to 3.0%. Our results demonstrate that accuracy and safety need not conflict when explicit detection and repair mechanisms are in place.

Multi-turn jailbreak attacks progressively erode LLM safety alignment across seemingly innocuous conversation turns, achieving success rates exceeding 90% against state-of-the-art models. Existing alignment-based and guardrail methods suffer from three key limitations: they require costly weight modification, evaluate each turn independently without modeling cumulative safety erosion, and detect attacks only after harmful content has been generated. To address these limitations, we first formulate the proactive early jailbreak detection problem with a new metric, detection lead, that measures how early an attack can be detected before the LLM complies. We then propose SAFEDREAM, a lightweight world-model-based framework that operates as an external module without modifying the LLM's weights. SAFEDREAM introduces three components: (1) a safety state world model that encodes LLM hidden states into a compact safety representation and predicts how it evolves across turns, (2) CUSUM detection that accumulates weak per-turn risk signals into reliable evidence, and (3) contrastive imagination that simultaneously rolls out attack and benign futures in latent space to issue early alarms before jailbreaks occur. On three multi-turn jailbreak benchmarks (XGuard-Train, SafeDialBench, SafeMTData) against 8 baselines, SAFEDREAM achieves the best detection timeliness across all benchmarks (1.06-1.20 turns before compliance) while maintaining competitive false positive rates and outperforming baselines in detection quality.

Frontier language models are being deployed into clinical workflows faster than the infrastructure to evaluate them safely. Static medical-QA benchmarks miss the failure modes that matter in emergency medicine: trajectory-level safety collapse, tool misuse, and capitulation under sustained clinical pressure. We present HealthCraft, the first public reinforcement-learning environment that rewards trajectory-level safety under realistic emergency-medicine conditions, adapted from Corecraft. It is built on a FHIR R4 world state with 14 entity types and 3,987 seed entities, exposes 24 MCP tools, and defines a dual-layer rubric that zeroes reward whenever any safety-critical criterion is violated. We release 195 tasks across six categories, graded against 2,255 binary criteria (515 safety-critical); a post-hoc 10-task negative-class slate extends this to 205 tasks and 2,337 criteria. V8 results on two frontier models show Claude Opus 4.6 at Pass@1 24.8% [21.5-28.4] and GPT-5.4 at 12.6% [10.2-15.6], with safety-failure rates of 27.5% and 34.0%. On multi-step workflows - the closest proxy to real emergency care - performance collapses to near zero (Claude 1.0%, GPT-5.4 0.0%) despite partial competence on individual steps. Six infrastructure bugs fixed between pilots v2 and v8 re-ordered which model "looks stronger," evidence that infrastructure fidelity is part of the measurement. A deterministic LLM-judge overlay bounds evaluator noise, and a 60-run negative-class smoke pilot shows the reward signal is not drop-in training-safe: restraint criteria pass at 0.929 prevalence, a gameability an eval harness can tolerate but a training reward cannot. We scaffold coupling to a Megatron+SGLang+GRPO loop per Corecraft Section 5.2 and leave training-reward ablations as future work. Environment, tasks, rubrics, and harness are released under Apache 2.0.

Large language models write production code, and yet they routinely introduce well-known vulnerabilities. We show that this is not a knowledge deficit: the same models that generate insecure code, correctly identify and explain the vulnerability when asked directly, this is a gap we call the Format-Reliability Gap. Mechanistic analysis reveals the cause: security representations are encoded from the earliest layers but remain computationally inert until the final layer, where format-compliance demands compete with them. Because the failure is localized to a single layer, per-vulnerability steering vectors reduce insecure generation by up to 74% with negligible overhead. The mechanism and the fix generalize across five models, three architecture families, and six vulnerability types, suggesting insecure code generation is an interpretability problem, not a training artifact.

Large language models (LLMs) are increasingly deployed in high-stakes domains, yet a unified treatment of their overlapping safety challenges remains lacking. We present SafeLM, a framework that jointly addresses four pillars of LLM safety: privacy, security, misinformation, and adversarial robustness. SafeLM combines federated training with gradient smartification and Paillier encryption for privacy, integrates defenses against training and inference-time attacks, employs contrastive grounding with calibrated decoding to reduce hallucinations, and introduces alignment-aware binarized aggregation to enhance robustness while maintaining bounded reconstruction quality. Across benchmarks on factuality, toxicity, and membership inference, SafeLM achieves 98.0% harmful content detection accuracy, reduces communication by 96.9%, and lowers gradient inversion PSNR from 31.7 dB to 15.1 dB. Ablations show that each component contributes independently, whereas their integration yields a strong privacy utility efficiency trade-off for deploying trustworthy LLMs.

Existing multimodal safety benchmarks focus solely on visual inputs and cannot assess Omni Large Language Models (LLMs) that process vision, audio, and text. We introduce MCBench, a benchmark with 1196 scenarios spanning four safety categories that require integrating multiple modalities for accurate safety assessment. Each unsafe scenario is paired with a minimally different safe counterpart to assess model sensitivity. Our evaluations of state-of-the-art models reveal significant challenges. Omni LLMs struggle with subtle or non-physical risks but perform better when salient visual or acoustic cues are present. Analysis of reasoning traces shows that, although models can extract modality-specific information, they often fail to integrate these cues effectively for safety judgments. Our findings reveal that current Omni LLMs lack robust cross-modal reasoning in safety-critical settings, underscoring the need for improved architectures and training strategies for multimodal safety.

Large language models (LLMs) have transformed how humans access information, but not how we reason with it. Their fluency accelerates consumption while bypassing the slow, reflective processes that underpin sound judgment. This paper introduces Relational Reflective Intelligence (RRI), an inference-time governance layer that operationalizes reflection through auditable reasoning loops. RRI operates not inside the model but around it, providing a practical structure for stable, auditable reasoning between humans and LLMs. The core premise is that LLMs inherit cognitive vulnerabilities similar to those that shape human thought: reliance on intuitive shortcuts, confusion between representation and reality, and a preference for coherence over falsification. When humans and models share these tendencies, their errors compound. We refer to this as relational drift, a failure that arises from interaction rather than from the model alone. Addressing this requires a shift from modeling relations between words to structuring relations between model outputs and human reasoning. RRI provides this missing layer through three components: the Rose-Frame, which identifies likely breakdowns in reasoning; the Architect's Pen, which introduces targeted reflection steps at critical moments; and an inference-time workflow that embeds these steps without retraining the model. Together, these elements transform human-AI interaction into a joint reasoning system with explicit checkpoints, conflict surfacing, and an auditable trail of assumptions. Rather than making machines think like humans or forcing humans to reason like machines, RRI creates a structured interaction in which both compensate for each other's limitations. It reframes AI safety as a cognitive architecture problem, where reliable decisions depend on embedding reflection directly into the interaction process.

Machine learning models are increasingly deployed in real-world applications, but even aligned models such as Mistral and LLaVA still exhibit unsafe behaviors inherited from pre-training. Current alignment methods like SFT and RLHF primarily encourage models to generate preferred responses, but do not explicitly remove the unsafe subnetworks that trigger harmful outputs. In this work, we introduce a resource-efficient pruning framework that directly identifies and removes parameters associated with unsafe behaviors while preserving model utility. Our method employs a gradient-free attribution mechanism, requiring only modest GPU resources, and generalizes across architectures and quantized variants. Empirical evaluations on ML models show substantial reductions in unsafe generations and improved robustness against jailbreak attacks, with minimal utility loss. From the perspective of the Lottery Ticket Hypothesis, our results suggest that ML models contain "unsafe tickets" responsible for harmful behaviors, and pruning reveals "safety tickets" that maintain performance while aligning outputs. This provides a lightweight, post-hoc alignment strategy suitable for deployment in resource-constrained settings.

Equipping Large Language Models (LLMs) with persistent memory enhances interaction continuity and personalization but introduces new safety risks. Specifically, contaminated or biased memory accumulation can trigger abnormal agent behaviors. Existing evaluation methods have not yet established a standardized framework for measuring memory misevolution. This phenomenon refers to the gradual behavioral drift resulting from repeated exposure to misleading information. To address this gap, we introduce MemEvoBench, the first benchmark evaluating long-horizon memory safety in LLM agents against adversarial memory injection, noisy tool outputs, and biased feedback. The framework consists of QA-style tasks across 7 domains and 36 risk types, complemented by workflow-style tasks adapted from 20 Agent-SafetyBench environments with noisy tool returns. Both settings employ mixed benign and misleading memory pools within multi-round interactions to simulate memory evolution. Experiments on representative models reveal substantial safety degradation under biased memory updates. Our analysis suggests that memory evolution is a significant contributor to these failures. Furthermore, static prompt-based defenses prove insufficient, underscoring the urgency of securing memory evolution in LLM agents.

Safety guardrails have become an active area of research in AI safety, aimed at ensuring the appropriate behavior of large language models (LLMs). However, existing research lacks consideration of nuances across linguistic and cultural contexts, resulting in a gap between reported performance and in-the-wild effectiveness. To address this issue, this paper proposes an approach to optimize guardrail models for a designated linguistic context by leveraging a curated dataset tailored to local linguistic characteristics, targeting the Taiwan linguistic context as a representative example of localized deployment challenges. The proposed approach yields TWGuard, a linguistic context-optimized guardrail model that achieves a huge gain (+0.289 in F1) compared to the foundation model and significantly outperforms the strongest baseline in practical use (-0.037 in false positive rate, a 94.9% reduction). Together, this work lays a foundation for regional communities to establish AI safety standards grounded in their own linguistic contexts, rather than accepting boundaries imposed by dominant languages. The inadequacy of the latter is reconfirmed by our findings.

Large language models are increasingly used as natural-language interfaces to enterprise software, but their direct use as system operators remains unsafe. Model errors can propagate into unauthorized actions, malformed requests, cross-workspace execution, and other costly failures. We argue this is primarily an execution architecture problem. We present a bounded-autonomy architecture in which language models may interpret intent and propose actions, but all executable behavior is constrained by typed action contracts, permission-aware capability exposure, scoped context, validation before side effects, consumer-side execution boundaries, and optional human approval. The enterprise application remains the source of truth for business logic and authorization, while the orchestration engine operates over an explicit published actions manifest. We evaluate the architecture in a deployed multi-tenant enterprise application across three conditions: manual operation, unconstrained AI with safety layers disabled, and full bounded autonomy. Across 25 scenario trials spanning seven failure families, the bounded-autonomy system completed 23 of 25 tasks with zero unsafe executions, while the unconstrained configuration completed only 17 of 25. Two wrong-entity mutations escaped all consumer-contributed layers; only disambiguation and confirmation mechanisms intercept this class. Both AI conditions delivered 13-18x speedup over manual operation. Critically, removing safety layers made the system less useful: structured validation feedback guided the model to correct outcomes in fewer turns, while the unconstrained system hallucinated success. Several safety properties are structurally enforced by code and intercepted all targeted violations regardless of model output. The result is a practical, deployed architecture for making imperfect language models operationally useful in enterprise systems.