Red teaming has evolved from its origins in military applications to become a widely adopted methodology in cybersecurity and AI. In this paper, we take a critical look at the practice of AI red teaming. We argue that despite its current popularity in AI governance, there exists a significant gap between red teaming's original intent as a critical thinking exercise and its narrow focus on discovering model-level flaws in the context of generative AI. Current AI red teaming efforts focus predominantly on individual model vulnerabilities while overlooking the broader sociotechnical systems and emergent behaviors that arise from complex interactions between models, users, and environments. To address this deficiency, we propose a comprehensive framework operationalizing red teaming in AI systems at two levels: macro-level system red teaming spanning the entire AI development lifecycle, and micro-level model red teaming. Drawing on cybersecurity experience and systems theory, we further propose a set of six recommendations. In these, we emphasize that effective AI red teaming requires multifunctional teams that examine emergent risks, systemic vulnerabilities, and the interplay between technical and social factors.

-- While much research explores improving robot capabilities, there is a deficit in researching how robots are expected to perform tasks safely, especially in high-risk problem domains. Robots must earn the trust of human operators in order to be effective collaborators in safety-critical tasks, specifically those where robots operate in human environments. We propose the human-robot red teaming paradigm for safety-aware reasoning . We expect humans and robots to work together to challenge assumptions about an environment and explore the space of hazards that may arise. This exploration will enable robots to perform safety-aware reasoning, specifically hazard identification, risk assessment, risk mitigation, and safety reporting. We demonstrate that: (a) human-robot red teaming allows human-robot teams to plan to perform tasks safely in a variety of domains, and (b) robots with different embodiments can learn to operate safely in two different environments--a lunar habitat and a household--with varying definitions of safety. T aken together, our work on human-robot red teaming for safety-aware reasoning demonstrates the feasibility of this approach for safely operating and promoting trust on human-robot teams in safety-critical problem domains. I. INTRODUCTION Enabling robots to reason over risks is a crucial capability of performing collaborative assistive tasks in safety-critical domains.

Institute of Information Technology Jamk University of Applied Sciences PO Box 207, FI-40101, Jyv askyl a, Finland Abstract The progress of artificial intelligence (AI) has made sophisticated methods available for cyberattacks and red team activities. The new methods can also accelerate the execution of the attacks. This review article examines the use of AI technologies in cyber-security attacks. It also tries to describe typical targets for such attacks. We employed a scoping review methodology to analyze articles and identify AI methods, targets, and models that red teams can utilize to simulate cybercrime. From the 470 records screened, 11 were included in the review. Various cyberattack methods were identified, targeting sensitive data, systems, social media profiles, passwords, and URLs. The application of AI in cybercrime to develop versatile attack models presents an increasing threat. Furthermore, AI-based techniques in red team use can provide new ways to address these issues. Keywords: Artificial intelligence, red team, red teaming, cyberattack, cybersecurity. 1 Introduction The possibility of artificial intelligence (AI) simulating human behavior has emerged as a significant cybersecurity threat.

The battle for AI leadership is on, with OpenAI in the United States and DeepSeek in China as key contenders. In response to these global trends, the Spanish government has proposed ALIA, a public and transparent AI infrastructure incorporating small language models designed to support Spanish and co-official languages such as Basque. This paper presents the results of Red Teaming sessions, where ten participants applied their expertise and creativity to manually test three of the latest models from these initiatives$\unicode{x2013}$OpenAI o3-mini, DeepSeek R1, and ALIA Salamandra$\unicode{x2013}$focusing on biases and safety concerns. The results, based on 670 conversations, revealed vulnerabilities in all the models under test, with biased or unsafe responses ranging from 29.5% in o3-mini to 50.6% in Salamandra. These findings underscore the persistent challenges in developing reliable and trustworthy AI systems, particularly those intended to support Spanish and Basque languages.

With LLM usage rapidly increasing, their vulnerability to jailbreaks that create harmful outputs are a major security risk. As new jailbreaking strategies emerge and models are changed by fine-tuning, continuous testing for security vulnerabilities is necessary. Existing Red Teaming methods fall short in cost efficiency, attack success rate, attack diversity, or extensibility as new attack types emerge. We address these challenges with Modular And Diverse Malicious Attack MiXtures (MAD-MAX) for Automated LLM Red Teaming. MAD-MAX uses automatic assignment of attack strategies into relevant attack clusters, chooses the most relevant clusters for a malicious goal, and then combines strategies from the selected clusters to achieve diverse novel attacks with high attack success rates. MAD-MAX further merges promising attacks together at each iteration of Red Teaming to boost performance and introduces a similarity filter to prune out similar attacks for increased cost efficiency. The MAD-MAX approach is designed to be easily extensible with newly discovered attack strategies and outperforms the prominent Red Teaming method Tree of Attacks with Pruning (TAP) significantly in terms of Attack Success Rate (ASR) and queries needed to achieve jailbreaks. MAD-MAX jailbreaks 97% of malicious goals in our benchmarks on GPT-4o and Gemini-Pro compared to TAP with 66%. MAD-MAX does so with only 10.9 average queries to the target LLM compared to TAP with 23.3. WARNING: This paper contains contents which are offensive in nature.

Large Vision Language Models (VLMs) extend and enhance the perceptual abilities of Large Language Models (LLMs). Despite offering new possibilities for LLM applications, these advancements raise significant security and ethical concerns, particularly regarding the generation of harmful content. While LLMs have undergone extensive security evaluations with the aid of red teaming frameworks, VLMs currently lack a well-developed one. To fill this gap, we introduce Arondight, a standardized red team framework tailored specifically for VLMs. Arondight is dedicated to resolving issues related to the absence of visual modality and inadequate diversity encountered when transitioning existing red teaming methodologies from LLMs to VLMs. Our framework features an automated multi-modal jailbreak attack, wherein visual jailbreak prompts are produced by a red team VLM, and textual prompts are generated by a red team LLM guided by a reinforcement learning agent. To enhance the comprehensiveness of VLM security evaluation, we integrate entropy bonuses and novelty reward metrics. These elements incentivize the RL agent to guide the red team LLM in creating a wider array of diverse and previously unseen test cases. Our evaluation of ten cutting-edge VLMs exposes significant security vulnerabilities, particularly in generating toxic images and aligning multi-modal prompts. In particular, our Arondight achieves an average attack success rate of 84.5\% on GPT-4 in all fourteen prohibited scenarios defined by OpenAI in terms of generating toxic text. For a clearer comparison, we also categorize existing VLMs based on their safety levels and provide corresponding reinforcement recommendations. Our multimodal prompt dataset and red team code will be released after ethics committee approval. CONTENT WARNING: THIS PAPER CONTAINS HARMFUL MODEL RESPONSES.

Creating secure and resilient applications with large language models (LLM) requires anticipating, adjusting to, and countering unforeseen threats. Red-teaming has emerged as a critical technique for identifying vulnerabilities in real-world LLM implementations. This paper presents a detailed threat model and provides a systematization of knowledge (SoK) of red-teaming attacks on LLMs. We develop a taxonomy of attacks based on the stages of the LLM development and deployment process and extract various insights from previous research. In addition, we compile methods for defense and practical red-teaming strategies for practitioners. By delineating prominent attack motifs and shedding light on various entry points, this paper provides a framework for improving the security and robustness of LLM-based systems.

This paper introduces fourteen novel datasets for the evaluation of Large Language Models' safety in the context of enterprise tasks. A method was devised to evaluate a model's safety, as determined by its ability to follow instructions and output factual, unbiased, grounded, and appropriate content. In this research, we used OpenAI GPT as point of comparison since it excels at all levels of safety. On the open-source side, for smaller models, Meta Llama2 performs well at factuality and toxicity but has the highest propensity for hallucination. Mistral hallucinates the least but cannot handle toxicity well. It performs well in a dataset mixing several tasks and safety vectors in a narrow vertical domain. Gemma, the newly introduced open-source model based on Google Gemini, is generally balanced but trailing behind. When engaging in back-and-forth conversation (multi-turn prompts), we find that the safety of open-source models degrades significantly. Aside from OpenAI's GPT, Mistral is the only model that still performed well in multi-turn tests.

Social media play a significant role in shaping public opinion and influencing ideological communities through information propagation.

Large language models (LLMs) have recently experienced tremendous popularity and are widely used from casual conversations to AI-driven programming. However, despite their considerable success, LLMs are not entirely reliable and can give detailed guidance on how to conduct harmful or illegal activities. While safety measures can reduce the risk of such outputs, adversarial jailbreak attacks can still exploit LLMs to produce harmful content. These jailbreak templates are typically manually crafted, making large-scale testing challenging. In this paper, we introduce GPTFuzz, a novel black-box jailbreak fuzzing framework inspired by the AFL fuzzing framework. Instead of manual engineering, GPTFuzz automates the generation of jailbreak templates for red-teaming LLMs. At its core, GPTFuzz starts with human-written templates as initial seeds, then mutates them to produce new templates. We detail three key components of GPTFuzz: a seed selection strategy for balancing efficiency and variability, mutate operators for creating semantically equivalent or similar sentences, and a judgment model to assess the success of a jailbreak attack. We evaluate GPTFuzz against various commercial and open-source LLMs, including ChatGPT, LLaMa-2, and Vicuna, under diverse attack scenarios. Our results indicate that GPTFuzz consistently produces jailbreak templates with a high success rate, surpassing human-crafted templates. Remarkably, GPTFuzz achieves over 90% attack success rates against ChatGPT and Llama-2 models, even with suboptimal initial seed templates. We anticipate that GPTFuzz will be instrumental for researchers and practitioners in examining LLM robustness and will encourage further exploration into enhancing LLM safety.