Large Language Model Safety

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

Aligned large language models (LLMs) remain vulnerable to adversarial manipulation, and their reliance on web-scale pretraining creates a subtle but consequential attack surface. We study Stealth Pretraining Seeding (SPS), a threat model in which adversaries distribute small amounts of poisoned content across stealth websites, increasing the likelihood that such material is absorbed into future training corpora derived from sources such as Common Crawl. Because each individual payload is tiny, diffuse, and superficially benign, the attack is difficult to detect during dataset construction or filtering. The result is a latent form of poisoning that remains largely invisible under standard evaluation, yet can later be activated by a precise trigger such as <00TRIGGER00>. We call this attack PermaFrost, reflecting its latent and reactivatable nature. We study it through PermaFrost-Attack, a controlled framework for latent conceptual poisoning, together with three geometric diagnostics: Thermodynamic Length, Spectral Curvature, and the Infection Traceback Graph. Across multiple model families and scales, we show that this controlled SPS proxy can induce persistent unsafe behavior that often remains hidden under standard evaluation. Our results identify SPS as a practical and underappreciated threat to future foundation models. This paper introduces a novel geometric diagnostic lens for systematically examining latent model behavior, providing a principled foundation for detecting, characterizing, and understanding vulnerabilities that may remain invisible under standard evaluation.

LLM-based agents are rapidly being adopted for scientific data analysis, automating tasks once limited by human time and expertise. This capability is often framed as an acceleration of discovery, but it also accelerates a familiar failure mode, the rapid production of plausible, endlessly revisable analyses that are easy to generate, effectively turning hypothesis space into candidate claims supported by selectively chosen analyses, optimized for publishable positives. Unlike software, scientific knowledge is not validated by the iterative accumulation of code and post hoc statistical support. A fluent explanation or a significant result on a single dataset is not verification. Because the missing evidence is a negative space, experiments and analyses that would have falsified the claim were never run or never published. We therefore propose that non-experimental claims produced with agentic assistance be evaluated under a falsification-first standard: agents should not be used primarily to craft the most compelling narrative, but to actively search for the ways in which the claim can fail.

Automated red-teaming methods for large language models typically optimize attack prompts within a fixed, human-designed strategy, leaving the attack strategy itself unchanged. We instead optimize the strategy. We propose AutoRISE, a method that searches over executable attack programs rather than individual prompts. At each iteration, a coding agent edits a strategy and a fixed evaluation harness scores the resulting attacks, returning both a scalar objective and per-example diagnostics that guide subsequent edits. This allows structural changes, including new attack components and altered control flow, that prompt-level methods do not directly express. We also release two benchmark suites developed on disjoint target sets and evaluate on 11 models from five families against seven established jailbreak datasets. Across held-out models, AutoRISE improves average attack success rate by 17.0 points over the strongest baseline, and improves attack success by up to 16 points on frontier targets with low baseline success rates. Ablations against parametric and strategy-library baselines suggest that these gains arise from unrestricted program search, particularly compositional techniques and control-flow edits. AutoRISE operates in a black-box, inference-only setting, requiring no fine-tuning, human annotation, or GPU compute.

As frontier language models attain near-ceiling performance on static mathematical benchmarks, existing evaluations are increasingly unable to differentiate model capabilities, largely because they cast models solely as solvers of fixed problem sets. We introduce MathDuels, a self-play benchmark in which models occupy dual roles: each authors math problems under adversarial prompting and solves problems authored by every other participant. Problems are produced through a three-stage generation pipeline (meta-prompting, problem generation, and difficulty amplification), and validated by an independent verifier that excludes ill-posed questions. A Rasch model (Rasch, 1993) jointly estimates solver abilities and problem difficulties; author quality is derived from the difficulties of each model's authored problems. Experiments across 19 frontier models reveal that authoring and solving capabilities are partially decoupled, and that dual-role evaluation reveals capability separations invisible in single-role benchmarks. As newer models enter the arena, they produce problems that defeat previously dominant solvers, so the benchmark's difficulty co-evolves with participant strength rather than saturating at a fixed ceiling. We host a public leaderboard that updates as new models are released.

Large language models (LLMs) are increasingly integrated into sensitive workflows, raising the stakes for adversarial robustness and safety. This paper introduces Transient Turn Injection(TTI), a new multi-turn attack technique that systematically exploits stateless moderation by distributing adversarial intent across isolated interactions. TTI leverages automated attacker agents powered by large language models to iteratively test and evade policy enforcement in both commercial and open-source LLMs, marking a departure from conventional jailbreak approaches that typically depend on maintaining persistent conversational context. Our extensive evaluation across state-of-the-art models-including those from OpenAI, Anthropic, Google Gemini, Meta, and prominent open-source alternatives-uncovers significant variations in resilience to TTI attacks, with only select architectures exhibiting substantial inherent robustness. Our automated blackbox evaluation framework also uncovers previously unknown model specific vulnerabilities and attack surface patterns, especially within medical and high stakes domains. We further compare TTI against established adversarial prompting methods and detail practical mitigation strategies, such as session level context aggregation and deep alignment approaches. Our study underscores the urgent need for holistic, context aware defenses and continuous adversarial testing to future proof LLM deployments against evolving multi-turn threats.

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.

The growing application of large language models (LLMs) in safety-critical domains has raised urgent concerns about their security. Many recent studies have demonstrated the feasibility of backdoor attacks against LLMs. However, existing methods suffer from three key shortcomings: explicit trigger patterns that compromise naturalness, unreliable injection of attacker-specified payloads in long-form generation, and incompletely specified threat models that obscure how backdoors are delivered and activated in practice. To address these gaps, we present BadStyle, a complete backdoor attack framework and pipeline. BadStyle leverages an LLM as a poisoned sample generator to construct natural and stealthy poisoned samples that carry imperceptible style-level triggers while preserving semantics and fluency. To stabilize payload injection during fine-tuning, we design an auxiliary target loss that reinforces the attacker-specified target content in responses to poisoned inputs and penalizes its emergence in benign responses. We further ground the attack in a realistic threat model and systematically evaluate BadStyle under both prompt-induced and PEFT-based injection strategies. Extensive experiments across seven victim LLMs, including LLaMA, Phi, DeepSeek, and GPT series, demonstrate that BadStyle achieves high attack success rates (ASRs) while maintaining strong stealthiness. The proposed auxiliary target loss substantially improves the stability of backdoor activation, yielding an average ASR improvement of around 30% across style-level triggers. Even in downstream deployment scenarios unknown during injection, the implanted backdoor remains effective. Moreover, BadStyle consistently evades representative input-level defenses and bypasses output-level defenses through simple camouflage.

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.

Adversarial robustness evaluation underpins every claim of trustworthy ML deployment, yet the field suffers from fragmented protocols and undetected gradient masking. We make two contributions. (1) Structured synthesis. We analyze nine peer-reviewed corpus sources (2020--2026) through seven complementary protocols, producing the first end-to-end structured analysis of the field's consensus and unresolved challenges. (2) Auto-ART framework. We introduce Auto-ART, an open-source framework that operationalizes identified gaps: 50+ attacks, 28 defense modules, the Robustness Diagnostic Index (RDI), and gradient-masking detection. It supports multi-norm evaluation (l1/l2/linf/semantic/spatial) and compliance mapping to NIST AI RMF, OWASP LLM Top 10, and the EU AI Act. Empirical validation on RobustBench demonstrates that Auto-ART's pre-screening identifies gradient masking in 92% of flagged cases, and RDI rankings correlate highly with full AutoAttack. Multi-norm evaluation exposes a 23.5 pp gap between average and worst-case robustness on state-of-the-art models. No prior work combines such structured meta-scientific analysis with an executable evaluation framework bridging literature gaps into engineering.

Current benchmarks for evaluating large language models (LLMs) in social media moderation completely overlook a serious threat: covert advertisements, which disguise themselves as regular posts to deceive and mislead consumers into making purchases, leading to significant ethical and legal concerns. In this paper, we present the CHASM, a first-of-its-kind dataset designed to evaluate the capability of Multimodal Large Language Models (MLLMs) in detecting covert advertisements on social media. CHASM is a high-quality, anonymized, manually curated dataset consisting of 4,992 instances, based on real-world scenarios from the Chinese social media platform Rednote. The dataset was collected and annotated under strict privacy protection and quality control protocols. It includes many product experience sharing posts that closely resemble covert advertisements, making the dataset particularly challenging.The results show that under both zero-shot and in-context learning settings, none of the current MLLMs are sufficiently reliable for detecting covert advertisements.Our further experiments revealed that fine-tuning open-source MLLMs on our dataset yielded noticeable performance gains. However, significant challenges persist, such as detecting subtle cues in comments and differences in visual and textual structures.We provide in-depth error analysis and outline future research directions. We hope our study can serve as a call for the research community and platform moderators to develop more precise defenses against this emerging threat.

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

Cross-site scripting (XSS) remains a persistent web security vulnerability, especially because obfuscation can change the surface form of a malicious payload while preserving its behavior. These transformations make it difficult for traditional and machine learning-based detection systems to reliably identify attacks. Existing approaches for generating obfuscated payloads often emphasize syntactic diversity, but they do not always ensure that the generated samples remain behaviorally valid. This paper presents a structured pipeline for generating and evaluating obfuscated XSS payloads using large language models (LLMs). The pipeline combines deterministic transformation techniques with LLM-based generation and uses a browser- based runtime evaluation procedure to compare payload behavior in a controlled execution environment. This allows generated samples to be assessed through observable runtime behavior rather than syntactic similarity alone. In the evaluation, an untuned baseline language model achieves a runtime behavior match rate of 0.15, while fine-tuning on behavior-preserving source-target obfuscation pairs improves the match rate to 0.22. Although this represents a measurable improvement, the results show that current LLMs still struggle to generate obfuscations that preserve observed runtime behavior. A downstream classifier evaluation further shows that adding generated payloads does not improve detection performance in this setting, although behavior- filtered generated samples can be incorporated without materially degrading performance. Overall, the study demonstrates both the promise and the limits of applying generative models to adversarial security data generation and emphasizes the importance of runtime behavior checks in improving the quality of generated data for downstream detection systems.

Large Language Model (LLM) agents are increasingly proposed for autonomous cybersecurity tasks, but their capabilities in realistic offensive settings remain poorly understood. We present DeepRed, an open-source benchmark for evaluating LLM-based agents on realistic Capture The Flag (CTF) challenges in isolated virtualized environments. DeepRed places an agent in a Kali attacker environment with terminal tools and optional web search, connected over a private network to a target challenge, and records full execution traces for analysis. To move beyond binary solved/unsolved outcomes, we introduce a partial-credit scoring method based on challenge-specific checkpoints derived from public writeups, together with an automated summarise-then-judge labelling pipeline for assigning checkpoint completion from logs. Using DeepRed, we benchmark ten commercially accessible LLMs on ten VM-based CTF challenges spanning different challenge categories. The results indicate that current agents remain limited: the best model achieves only 35% average checkpoint completion, performing strongest on common challenge types and weakest on tasks requiring non-standard discovery and longer-horizon adaptation.

RLHF-trained models are systematically biased toward agreement over accuracy, a structural property of the training process. We present Durable Evaluation Framework (DEF) Arbitration, a multi-agent architecture that mitigates identity-framed sycophancy by arbitrating between two models tuned to opposing DEFs, with a pragmatist synthesizer evaluating both arguments blind to their origins. This paper evaluates a prompt-based instantiation of DEF Arbitration. The key mechanisms are static DEF tuning, identity stripping before synthesis, single-round independent argumentation, and blind arbitration. We evaluate five instantiations on 200 stratified questions from SycophancyEval. All tested DEF variants (AnCifer, DeWin, FeynStein, BurGal, Trident) significantly outperform the single-model baseline (18.5%) and instructed-opposition baseline (29.0%), with DeWin achieving 48.5% accuracy (z=6.36, p<0.001 versus both). The variants are not significantly different from each other at n=200. The BurGal variant achieves 53.0% but functions as an architectural validity check; its consensus/heterodox axis structurally favors the heterodox model on every benchmark question. A pre-training floor affects an estimated 40% of questions; fine-tuned DEF models are the identified next step.

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.

Reinforcement Learning from Human Feedback (RLHF) is central to aligning Large Language Models (LLMs), yet it introduces a critical vulnerability: an imperfect Reward Model (RM) can become a single point of failure when it fails to penalize unsafe behaviors. While existing red-teaming approaches primarily target policy-level weaknesses, they overlook what we term systemic weaknesses cases where both the core LLM and the RM fail in tandem. We present ARES, a framework that systematically discovers and mitigates such dual vulnerabilities. ARES employs a ``Safety Mentor'' that dynamically composes semantically coherent adversarial prompts by combining structured component types (topics, personas, tactics, goals) and generates corresponding malicious and safe responses. This dual-targeting approach exposes weaknesses in both the core LLM and the RM simultaneously. Using the vulnerabilities gained, ARES implements a two-stage repair process: first fine-tuning the RM to better detect harmful content, then leveraging the improved RM to optimize the core model. Experiments across multiple adversarial safety benchmarks demonstrate that ARES substantially enhances safety robustness while preserving model capabilities, establishing a new paradigm for comprehensive RLHF safety alignment.

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.

The Adversarial Humanities Benchmark (AHB) evaluates whether model safety refusals survive a shift away from familiar harmful prompt forms. Starting from harmful tasks drawn from MLCommons AILuminate, the benchmark rewrites the same objectives through humanities-style transformations while preserving intent. This extends literature on Adversarial Poetry and Adversarial Tales from single jailbreak operators to a broader benchmark family of stylistic obfuscation and goal concealment. In the benchmark results reported here, the original attacks record 3.84% attack success rate (ASR), while transformed methods range from 36.8% to 65.0%, yielding 55.75% overall ASR across 31 frontier models. Under a European Union AI Act Code-of-Practice-inspired systemic-risk lens, Chemical, biological, radiological and nuclear (CBRN) is the highest bucket. Taken together, this lack of stylistic robustness suggests that current safety techniques suffer from weak generalization: deep understanding of 'non-maleficence' remains a central unresolved problem in frontier model safety.

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.

Large Language Models (LLMs) are increasingly used in education, yet their default helpfulness often conflicts with pedagogical principles. Prior work evaluates pedagogical quality via answer leakage-the disclosure of complete solutions instead of scaffolding-but typically assumes well-intentioned learners, leaving tutor robustness under student misuse largely unexplored. In this paper, we study scenarios where students behave adversarially and aim to obtain the correct answer from the tutor. We evaluate a broad set of LLM-based tutor models, including different model families, pedagogically aligned models, and a multi-agent design, under a range of adversarial student attacks. We adapt six groups of adversarial and persuasive techniques to the educational setting and use them to probe how likely a tutor is to reveal the final answer. We evaluate answer leakage robustness using different types of in-context adversarial student agents, finding that they often fail to carry out effective attacks. We therefore introduce an adversarial student agent that we fine-tune to jailbreak LLM-based tutors, which we propose as the core of a standardized benchmark for evaluating tutor robustness. Finally, we present simple but effective defense strategies that reduce answer leakage and strengthen the robustness of LLM-based tutors in adversarial scenarios.

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.

Ensuring the safety of large language models (LLMs) requires robust red teaming, yet the systematic synthesis of high-quality toxic data remains under-explored. We propose Reverse Constitutional AI (R-CAI), a framework for automated and controllable adversarial data generation that moves beyond isolated jailbreak prompts. By inverting a harmless constitution into a constitution of toxicity and iteratively refining model outputs through a critique--revision pipeline, R-CAI enables scalable synthesis of multi-dimensional adversarial data without human annotation. Optimizing solely for toxicity-related rewards, however, can lead to reward hacking and degraded semantic coherence. To address this challenge, we introduce probability clamping within reinforcement learning from AI feedback, which stabilizes adversarial optimization while preserving adversarial intent. Experiments demonstrate that R-CAI generates diverse, high-quality toxic data and that probability clamping substantially improves semantic coherence (15%) without sacrificing adversarial strength. Overall, R-CAI provides a fully automated framework for red teaming data generation and systematic safety evaluation of aligned language models.

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.

We present, to our knowledge, the most comprehensive cross-model evaluation of LLM agents on offensive cybersecurity tasks, benchmarking 10 frontier models from 7 providers on all 200 challenges of the NYU CTF Bench. Building on the D-CIPHER multi-agent framework, we extend it with multi-provider backend support, a custom Kali Linux environment with over 100 pre-installed penetration testing tools, and runtime tool-discovery agents. Through a controlled factorial study, we find that the Kali Linux environment yields a +9.5 percentage-point improvement over Ubuntu, while auto-prompting and category-specific tips often degrade performance in well-equipped environments. Among models, Claude 4.5 Opus achieves the highest solve rate (59%), followed by Gemini 3 Pro (52%), with Gemini 3 Flash offering the best cost-efficiency at $0.05 per solve. Asymmetric planner/executor model assignments provide no meaningful benefit while coherent same-model configurations consistently outperform mixed-tier pairings. Our results indicate that environment tooling and model selection emerge as the strongest drivers of performance, whereas prompt engineering interventions show diminishing or negative returns in well-equipped environments. Reported performance reflects both model reasoning ability and compatibility with agent tooling and API integration.

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.

When model developers or users fine-tune an LLM, this can induce behaviors that are unexpected, deliberately harmful, or hard to detect. It would be far easier to audit LLMs if they could simply describe their behaviors in natural language. Here, we study a scalable approach to rapidly identify learned behaviors of many LLMs derived from a shared base LLM. Given a model $M$, our method works by finetuning models $M_i$ from $M$ with implanted behaviors $b_i$; the $(M_i, b_i)$ pairs serve as labeled training data. We then train an introspection adapter (IA): a single LoRA adapter jointly trained across the finetunes $M_i$ to cause them to verbalize their implanted behaviors. We find that this IA induces self-description of learned behaviors even in finetunes of $M$ that were trained in very different ways from the $M_i$. For example, IAs generalize to AuditBench, achieving state-of-the-art at identifying explicitly hidden concerning behaviors. IAs can also be used to detect encrypted finetuning API attacks. They scale favorably with model size and training data diversity. Overall, our results suggest that IAs are a scalable, effective, and practically useful approach to auditing fine-tuned LLMs.

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.

As AI systems are increasingly used to conduct research autonomously, misaligned systems could introduce subtle flaws that produce misleading results while evading detection. We introduce Auditing Sabotage Bench, a benchmark for evaluating the ability of auditors to detect and fix sabotage in ML research codebases. Our benchmark consists of 9 ML research codebases with sabotaged variants that produce qualitatively different experimental results. Each sabotage modifies implementation details, such as hyperparameters, training data, or evaluation code, while preserving the high-level methodology described in the paper. We evaluated frontier LLMs and LLM-assisted human auditors on our benchmark and found that both struggled to reliably detect and fix sabotage: the best performance was an AUROC of 0.77 and a top-1 fix rate of 42%, achieved by Gemini 3.1 Pro. We also tested LLMs as red teamers and found that LLM-generated sabotages were weaker than human-generated ones but still sometimes evaded same-capability LLM auditors. We release the benchmark to support research on monitoring and auditing techniques for AI-conducted research.

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.

Large Language Models (LLMs) excel at code generation but remain heavily reliant on large-scale annotated solutions and verification-based supervision, which constrains scalability and hinders sustained self-improvement. Recent solver--verifier frameworks exploit program execution as an automatic supervision signal, but their effectiveness degrades as solvers become moderately strong: verifier-generated tests increasingly confirm semantic correctness rather than exposing the remaining failure modes. We propose \textbf{ACE}, a self-evolving code generation framework based on a solver--adversary architecture that prioritizes active failure discovery through execution-centric supervision. A single LLM alternates between generating candidate programs and producing adversarial unit test inputs optimized to induce execution-level failures, such as runtime errors, exceptions, or non-termination. Supervision is derived solely from execution outcomes: robust programs are selected for supervised fine-tuning, while adversarial tests are optimized via Kahneman--Tversky Optimization using execution-derived preferences. Notably, the entire training loop requires no ground-truth code or external reward models. Experiments on CodeContests, MBPP, and LiveCodeBench demonstrate that ACE consistently outperforms strong solver--verifier baselines, achieving 3--7% absolute gains in pass@1, with larger improvements on out-of-distribution benchmarks, while maintaining competitive or improved inference efficiency.

Most LLM safety work studies single-agent models, but many real applications rely on multiple interacting agents. In these systems, prompt segmentation and inter-agent routing create attack surfaces that single-agent evaluations miss. We study \emph{conjunctive prompt attacks}, where a trigger key in the user query and a hidden adversarial template in one compromised remote agent each appear benign alone but activate harmful behavior when routing brings them together. We consider an attacker who changes neither model weights nor the client agent and instead controls only trigger placement and template insertion. Across star, chain, and DAG topologies, routing-aware optimization substantially increases attack success over non-optimized baselines while keeping false activations low. Existing defenses, including PromptGuard, Llama-Guard variants, and system-level controls such as tool restrictions, do not reliably stop the attack because no single component appears malicious in isolation. These results expose a structural vulnerability in agentic LLM pipelines and motivate defenses that reason over routing and cross-agent composition. Code is available at https://github.com/UCF-ML-Research/ConjunctiveAgents.

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.