We address the problem of data scarcity in harmful text classification for guardrailing applications and introduce GRAID (Geometric and Reflective AI-Driven Data Augmentation), a novel pipeline that leverages Large Language Models (LLMs) for dataset augmentation. GRAID consists of two stages: (i) generation of geometrically controlled examples using a constrained LLM, and (ii) augmentation through a multi-agentic reflective process that promotes stylistic diversity and uncovers edge cases. This combination enables both reliable coverage of the input space and nuanced exploration of harmful content. Using two benchmark data sets, we demonstrate that augmenting a harmful text classification dataset with GRAID leads to significant improvements in downstream guardrail model performance.

AI Alignment, primarily in the form of Reinforcement Learning from Human Feedback (RLHF), has been a cornerstone of the post-training phase in developing Large Language Models (LLMs). It has also been a popular research topic across various disciplines beyond Computer Science, including Philosophy and Law, among others, highlighting the socio-technical challenges involved. Nonetheless, except for the computational techniques related to alignment, there has been limited focus on the broader picture: the scope of these processes, which primarily rely on the selected objectives (values), and the data collected and used to imprint such objectives into the models. This work aims to reveal how alignment is understood and applied in practice from a value-setting and data-centric perspective. For this purpose, we investigate and survey (`audit') publicly available documentation released by 6 LLM development initiatives by 5 leading organizations shaping this technology, focusing on proprietary (OpenAI's GPT, Anthropic's Claude, Google's Gemini) and open-weight (Meta's Llama, Google's Gemma, and Alibaba's Qwen) initiatives, all published in the last 3 years. The findings are documented in detail per initiative, while there is also an overall summary concerning different aspects, mainly from a value-setting and data-centric perspective. On the basis of our findings, we discuss a series of broader related concerns.

We analyzed nearly 460,000 AI model cards from Hugging Face to examine how developers report risks. From these, we extracted around 3,000 unique risk mentions and built the \emph{AI Model Risk Catalog}. We compared these with risks identified by researchers in the MIT Risk Repository and with real-world incidents from the AI Incident Database. Developers focused on technical issues like bias and safety, while researchers emphasized broader social impacts. Both groups paid little attention to fraud and manipulation, which are common harms arising from how people interact with AI. Our findings show the need for clearer, structured risk reporting that helps developers think about human-interaction and systemic risks early in the design process. The catalog and paper appendix are available at: https://social-dynamics.net/ai-risks/catalog.

This research paper examines, from a multidimensional perspective (cognitive, social, ethical, and philosophical), how AI is transforming human thought. It highlights a cognitive offloading effect: the externalization of mental functions to AI can reduce intellectual engagement and weaken critical thinking. On the social level, algorithmic personalization creates filter bubbles that limit the diversity of opinions and can lead to the homogenization of thought and polarization. This research also describes the mechanisms of algorithmic manipulation (exploitation of cognitive biases, automated disinformation, etc.) that amplify AI's power of influence. Finally, the question of potential artificial consciousness is discussed, along with its ethical implications. The report as a whole underscores the risks that AI poses to human intellectual autonomy and creativity, while proposing avenues (education, transparency, governance) to align AI development with the interests of humanity.

Social identity theory (SIT) and social categorization theory (SCT) are two facets of the social identity approach (SIA) to understanding social phenomena. SIT and SCT are models that describe and explain how people interact with one another socially, connecting the individual to the group through an understanding of underlying psychological mechanisms and intergroup behaviour. SIT, originally developed in the 1970s, and SCT, a later, more general offshoot, have been broadly applied to a range of social phenomena among people. The rise of increasingly social machines embedded in daily life has spurned efforts on understanding whether and how artificial agents can and do participate in SIA activities. As agents like social robots and chatbots powered by sophisticated large language models (LLMs) advance, understanding the real and potential roles of these technologies as social entities is crucial. Here, I provide a primer on SIA and extrapolate, through case studies and imagined examples, how SIT and SCT can apply to artificial social agents. I emphasize that not all human models and sub-theories will apply. I further argue that, given the emerging competence of these machines and our tendency to be taken in by them, we experts may need to don the hat of the uncanny killjoy, for our own good.

Tool-augmented large language models (LLMs) leverage external functions to extend their capabilities, but inaccurate function calls can lead to inefficiencies and increased costs.Existing methods address this challenge by fine-tuning LLMs or using demonstration-based prompting, yet they often suffer from high training overhead and fail to account for inconsistent demonstration samples, which misguide the model's invocation behavior. In this paper, we trained a behavior-aligned retriever (BAR), which provides behaviorally consistent demonstrations to help LLMs make more accurate tool-using decisions. To train the BAR, we construct a corpus including different function-calling behaviors, i.e., calling or non-calling.We use the contrastive learning framework to train the BAR with customized positive/negative pairs and a dual-negative contrastive loss, ensuring robust retrieval of behaviorally consistent examples.Experiments demonstrate that our approach significantly reduces erroneous function calls while maintaining high task performance, offering a cost-effective and efficient solution for tool-augmented LLMs.

Large Language Models (LLMs) have demonstrated remarkable capabilities in natural language processing (NLP), including text generation, translation, summarization, and code synthesis, as a consequence of which revolutionizing a wide range of AI applications [10, 56, 45]. Models such as OpenAI's ChatGPT series, Google's Gemini, and Anthropic's Claude have been widely deployed in commercial systems, including search engines, customer support, software development tools, and personal assistants [45, 55, 3]. However, as their capabilities grow, so do their attack surfaces and the potential for misuse [51, 77, 50]. While the scale and specific nature of these vulnerabilities are new, the fundamental challenge of ensuring that powerful AI systems operate safely and align with human intent is a longstanding concern in the AI community. Foundational work, such as the identification of concrete problems in AI safety long before the current LLM era, laid the groundwork for understanding issues like reward hacking and negative side effects that remain highly relevant today [1]. The susceptibility arises because the models are trained on vast, yet imperfectly curated, datasets containing potentially harmful content, and because they interact with users through open-ended prompts that can be manipulated [48, 17, 16]. Researchers and practitioners are increasingly concerned that these systems can be manipulated, misused, or even behave in misaligned and potentially deceptive ways [25, 42, 6]. Consequently, the security and alignment of LLMs have become critical areas of study, requiring an understanding of emergent threats and robust, multi-faceted defenses [17, 70, 43].

Multi-turn interactions with language models (LMs) pose critical safety risks, as harmful intent can be strategically spread across exchanges. Yet, the vast majority of prior work has focused on single-turn safety, while adaptability and diversity remain among the key challenges of multi-turn red-teaming. To address these challenges, we present X-Teaming, a scalable framework that systematically explores how seemingly harmless interactions escalate into harmful outcomes and generates corresponding attack scenarios. X-Teaming employs collaborative agents for planning, attack optimization, and verification, achieving state-of-the-art multi-turn jailbreak effectiveness and diversity with success rates up to 98.1% across representative leading open-weight and closed-source models. In particular, X-Teaming achieves a 96.2% attack success rate against the latest Claude 3.7 Sonnet model, which has been considered nearly immune to single-turn attacks. Building on X-Teaming, we introduce XGuard-Train, an open-source multi-turn safety training dataset that is 20x larger than the previous best resource, comprising 30K interactive jailbreaks, designed to enable robust multi-turn safety alignment for LMs. Our work offers essential tools and insights for mitigating sophisticated conversational attacks, advancing the multi-turn safety of LMs.

Elon Musk's xAI filed a lawsuit against Apple and OpenAI on Monday, accusing the companies of behaving like monopolies and claiming Apple deprioritized ChatGPT rivals like Grok in the App Store. "This is a tale of two monopolists joining forces to ensure their continued dominance in a world rapidly driven by the most powerful technology humanity has ever created: artificial intelligence," the lawsuit alleges. "Working in tandem, Defendants Apple and OpenAI have locked up markets to maintain their monopolies and prevent innovators like X and xAI from competing." Grok is currently ranked third in the App Store for free productivity apps--behind only ChatGPT and Gmail. The'uncensored' chatbot is also integrated into Musk's social platform X, which is the number one free news app in the App Store.

Paul McCartney enlisted a little help from artificial intelligence to complete the'last Beatles song' two years ago. The track, 'Now and Then', became the first Beatles music to reach number 1 in the UK for 64 years. Now, in an apparent effort to repeat its success, McCartney has once again used AI – on two more songs. The sophisticated tool called'MAL' is the creation of WingNut Films, the production company headed by Lord of the Rings director Peter Jackson. MAL has managed to extricate John Lennon's voice from two poor-quality demos he made shortly before his death.