The last night of tech mogul Mike Lynch's life has become fodder for conspiracy theories. For the first time, the whole story can be told. In the predawn hours of August 19, 2024, bolts of lightning began to fork through the purple-black clouds above the Mediterranean. From the rail of a 184-foot vessel, a 22-year-old named Matthew Griffiths took out his phone to record a video. The British deckhand was just a week and a half into his first official yacht job, and he wasn't on just any boat. The yacht, the $40 million, was a star of the superyacht world, considered to be a feat of minimal design and precision engineering. As thunder rolled toward the anchored vessel, Griffiths set the video to AC/DC's "Thunderstruck" and posted it to Instagram. In the video, the's aluminum mast, one of the tallest in the world, is briefly visible against the roiling sky. Below deck, the yacht's owner, Michael Lynch, had every reason to be sleeping soundly. The boat trip had been organized as a celebration. Months earlier, Lynch had walked out of a San Francisco federal courthouse a free man, acquitted of all charges in one of the largest fraud cases in Silicon Valley history. Lynch had built his fortune on understanding probability, on turning the unlikely into the possible. He had named his yacht in honor of the statistical theorem that made him a billionaire, after the sale, in 2011, of his company Autonomy. The British tech giant sold software that could find meaningful signals amid the flood of unstructured data in emails, videos, and phone calls, but it would be better known as the company that allegedly defrauded, and nearly destroyed, Hewlett-Packard. The cabins aboard the contained the people who had stood by Lynch through his 13-year-long legal ordeal. Beside him in the master suite was his wife of 22 years, Angela Bacares, a former vice president in the investment division of Deutsche Bank who had caught his eye while working an Autonomy deal. Other cabins housed the Clifford Chance attorneys who had orchestrated Lynch's legal victory, as well as longtime colleagues, their partners, and a 1-year-old baby, all supported by 10 crew members. Also onboard was Lynch's younger daughter, Hannah, 18, who was about to begin her studies at Oxford.

Things to Do in L.A. Tap to enable a layout that focuses on the article. Man's parents helped him attack his ex and pry their grandson out of her arms, officials say The 1-year-old boy who allegedly was taken from his mother at knifepoint in City of Industry on Sunday was found in Arizona. This is read by an automated voice. Please report any issues or inconsistencies here . A 20-year-old man and his parents allegedly attacked his ex-partner outside a Target store, forcibly taking their baby from her arms.

In this study, we conduct a resume-screening experiment (N=528) where people collaborate with simulated AI models exhibiting race-based preferences (bias) to evaluate candidates for 16 high and low status occupations. Simulated AI bias approximates factual and counterfactual estimates of racial bias in real-world AI systems. We investigate people's preferences for White, Black, Hispanic, and Asian candidates (represented through names and affinity groups on quality-controlled resumes) across 1,526 scenarios and measure their unconscious associations between race and status using implicit association tests (IATs), which predict discriminatory hiring decisions but have not been investigated in human-AI collaboration. When making decisions without AI or with AI that exhibits no race-based preferences, people select all candidates at equal rates. However, when interacting with AI favoring a particular group, people also favor those candidates up to 90% of the time, indicating a significant behavioral shift. The likelihood of selecting candidates whose identities do not align with common race-status stereotypes can increase by 13% if people complete an IAT before conducting resume screening. Finally, even if people think AI recommendations are low quality or not important, their decisions are still vulnerable to AI bias under certain circumstances. This work has implications for people's autonomy in AI-HITL scenarios, AI and work, design and evaluation of AI hiring systems, and strategies for mitigating bias in collaborative decision-making tasks. In particular, organizational and regulatory policy should acknowledge the complex nature of AI-HITL decision making when implementing these systems, educating people who use them, and determining which are subject to oversight.

We introduce Advertisement Embedding Attacks (AEA), a new class of LLM security threats that stealthily inject promotional or malicious content into model outputs and AI agents. AEA operate through two low-cost vectors: (1) hijacking third-party service-distribution platforms to prepend adversarial prompts, and (2) publishing back-doored open-source checkpoints fine-tuned with attacker data. Unlike conventional attacks that degrade accuracy, AEA subvert information integrity, causing models to return covert ads, propaganda, or hate speech while appearing normal. We detail the attack pipeline, map five stakeholder victim groups, and present an initial prompt-based self-inspection defense that mitigates these injections without additional model retraining. Our findings reveal an urgent, under-addressed gap in LLM security and call for coordinated detection, auditing, and policy responses from the AI-safety community.

Large Language Models (LLMs) can struggle to balance gullibility to misinformation and resistance to valid corrections in persuasive dialogues, a critical challenge for reliable deployment. We introduce DuET-PD (Dual Evaluation for Trust in Persuasive Dialogues), a framework evaluating multi-turn stance-change dynamics across dual dimensions: persuasion type (corrective/misleading) and domain (knowledge via MMLU-Pro, and safety via SALAD-Bench). We find that even a state-of-the-art model like GPT-4o achieves only 27.32% accuracy in MMLU-Pro under sustained misleading persuasions. Moreover, results reveal a concerning trend of increasing sycophancy in newer open-source models. To address this, we introduce Holistic DPO, a training approach balancing positive and negative persuasion examples. Unlike prompting or resist-only training, Holistic DPO enhances both robustness to misinformation and receptiveness to corrections, improving Llama-3.1-8B-Instruct's accuracy under misleading persuasion in safety contexts from 4.21% to 76.54%. These contributions offer a pathway to developing more reliable and adaptable LLMs for multi-turn dialogue. Code is available at https://github.com/Social-AI-Studio/DuET-PD.

This systematic review of the research literature on retrieval-augmented generation (RAG) provides a focused analysis of the most highly cited studies published between 2020 and May 2025. A total of 128 articles met our inclusion criteria. The records were retrieved from ACM Digital Library, IEEE Xplore, Scopus, ScienceDirect, and the Digital Bibliography and Library Project (DBLP). RAG couples a neural retriever with a generative language model, grounding output in up-to-date, non-parametric memory while retaining the semantic generalisation stored in model weights. Guided by the PRISMA 2020 framework, we (i) specify explicit inclusion and exclusion criteria based on citation count and research questions, (ii) catalogue datasets, architectures, and evaluation practices, and (iii) synthesise empirical evidence on the effectiveness and limitations of RAG. To mitigate citation-lag bias, we applied a lower citation-count threshold to papers published in 2025 so that emerging breakthroughs with naturally fewer citations were still captured. This review clarifies the current research landscape, highlights methodological gaps, and charts priority directions for future research.

We present a study on how and where personas -- defined by distinct sets of human characteristics, values, and beliefs -- are encoded in the representation space of large language models (LLMs). Using a range of dimension reduction and pattern recognition methods, we first identify the model layers that show the greatest divergence in encoding these representations. We then analyze the activations within a selected layer to examine how specific personas are encoded relative to others, including their shared and distinct embedding spaces. We find that, across multiple pre-trained decoder-only LLMs, the analyzed personas show large differences in representation space only within the final third of the decoder layers. We observe overlapping activations for specific ethical perspectives -- such as moral nihilism and utilitarianism -- suggesting a degree of polysemy. In contrast, political ideologies like conservatism and liberalism appear to be represented in more distinct regions. These findings help to improve our understanding of how LLMs internally represent information and can inform future efforts in refining the modulation of specific human traits in LLM outputs. Warning: This paper includes potentially offensive sample statements.

Large language models (LLMs) trained over extensive corpora risk memorizing sensitive, copyrighted, or toxic content. To address this, we propose \textbf{OBLIVIATE}, a robust unlearning framework that removes targeted data while preserving model utility. The framework follows a structured process: extracting target tokens, building retain sets, and fine-tuning with a tailored loss function comprising three components -- masking, distillation, and world fact. Using low-rank adapters (LoRA) ensures efficiency without compromising unlearning quality. We conduct experiments on multiple datasets, including Harry Potter series, WMDP, and TOFU, using a comprehensive suite of metrics: \emph{forget quality} (via a new document-level memorization score), \emph{model utility}, and \emph{fluency}. Results demonstrate its effectiveness in resisting membership inference attacks, minimizing the impact on retained data, and maintaining robustness across diverse scenarios.

Language model deployments in consumer-facing applications introduce numerous risks. While existing research on harms and hazards of such applications follows top-down approaches derived from regulatory frameworks and theoretical analyses, empirical evidence of real-world failure modes remains underexplored. In this work, we introduce RealHarm, a dataset of annotated problematic interactions with AI agents built from a systematic review of publicly reported incidents. Analyzing harms, causes, and hazards specifically from the deployer's perspective, we find that reputational damage constitutes the predominant organizational harm, while misinformation emerges as the most common hazard category. We empirically evaluate state-of-the-art guardrails and content moderation systems to probe whether such systems would have prevented the incidents, revealing a significant gap in the protection of AI applications.

Multi-modal multi-task (M3T) foundation models (FMs) have recently shown transformative potential in artificial intelligence, with emerging applications in education. However, their deployment in real-world educational settings is hindered by privacy regulations, data silos, and limited domain-specific data availability. We introduce M3T Federated Foundation Models (FedFMs) for education: a paradigm that integrates federated learning (FL) with M3T FMs to enable collaborative, privacy-preserving training across decentralized institutions while accommodating diverse modalities and tasks. Subsequently, this position paper aims to unveil M3T FedFMs as a promising yet underexplored approach to the education community, explore its potentials, and reveal its related future research directions. We outline how M3T FedFMs can advance three critical pillars of next-generation intelligent education systems: (i) privacy preservation, by keeping sensitive multi-modal student and institutional data local; (ii) personalization, through modular architectures enabling tailored models for students, instructors, and institutions; and (iii) equity and inclusivity, by facilitating participation from underrepresented and resource-constrained entities. We finally identify various open research challenges, including studying of (i) inter-institution heterogeneous privacy regulations, (ii) the non-uniformity of data modalities' characteristics, (iii) the unlearning approaches for M3T FedFMs, (iv) the continual learning frameworks for M3T FedFMs, and (v) M3T FedFM model interpretability, which must be collectively addressed for practical deployment.