Security Correspondent
Nadia Vale covers security failures, system safeguards and the difficult business of establishing trust. Her beat connects technical disclosures with their consequences for users and institutions. She favors specific threat models over sweeping claims that a system is safe or unsafe.
1,019 published articles · Showing 101–120
New research from arXiv CS. AI unveils significant advancements in detecting and mitigating fundamental vulnerabilities within Artificial Intelligence reasoning systems....
Read article →The foundational security of Artificial Intelligence systems is under renewed scrutiny, with a series of recent research papers exposing significant vulnerabilities across Graph Neural Networks (GNNs), Large Language Models (LLMs), and LLM-powered agents. These findings demonstra...
Read article →Recent research from arXiv CS. AI reveals significant advancements in AI model architecture and deployment, focusing on internal state decomposition for recurrent language models and sophisticated strategies for merging expert models without costly retraining....
Read article →Recent research published on arXiv CS. AI details a significant acceleration in AI's capacity to model and influence human social dynamics, revealing both advanced methods for social engineering and emergent defense strategies....
Read article →Recent research published on arXiv CS. AI reveals a critical pivot in AI development: a concerted effort to dismantle algorithmic opacity and establish verifiable accountability....
Read article →A ransomware group has publicly claimed responsibility for a breach at Foxconn, a global electronics manufacturing giant integral to the supply chains of major technology firms including Apple, Google, and Nvidia TechCrunch. This development underscores the persistent and escalat...
Read article →The digital battlefield extends beyond terrestrial networks. The recent US military wargame simulating a nuclear detonation in low-Earth orbit underscores the catastrophic fragility of orbital infrastructure, a vulnerability that parallels the complex attack surfaces of ambitious...
Read article →The latest advancements in AI for robotics and physical interaction, published in arXiv CS. AI on May 13, 2026, reveal a dual trajectory: enhanced perception for intelligent vehicles and more intuitive human-robot teleoperation....
Read article →A novel AI framework, AVA-DINO, has been introduced, aiming to advance zero-shot anomaly detection by fundamentally altering how machine learning models process normal and anomalous data. Developed as an anomaly-aware vision-language adaptation framework, AVA-DINO directly confro...
Read article →A recent research paper, arXiv:2605. 12069v1, introduces AVA-DINO, an anomaly-aware vision-language adaptation framework designed to enhance zero-shot anomaly detection....
Read article →LLMs, despite their advanced predictive capabilities, demonstrably struggle with fundamental causal inference, a critical vulnerability that undermines their reliability in high-stakes environments such as medicine, economics, and public policy. New research reveals current bench...
Read article →The latest advancements in Human Activity Recognition (HAR) leverage Large Language Model (LLM) backbones to achieve purportedly efficient and adaptive performance, a critical development given the technology's immediate deployment in sensitive applications such as precise gait a...
Read article →Recent deep learning research has unveiled a critical vulnerability in the current generation of self-supervised learning (SSL) methods for Automatic Modulation Classification (AMC). As detailed in a new arXiv publication, these methods often produce internal representations "ent...
Read article →Vision-Language-Action (VLA) models, critical for advanced autonomous systems, are demonstrably brittle in fine-grained manipulation, where even minor action errors during critical operational phases can rapidly escalate into irrecoverable failures arXiv CS. AI....
Read article →New research from arXiv CS. AI reveals fundamental security and fairness vulnerabilities within artificial intelligence deployments....
Read article →A new class of attack, dubbed AbO-DDoS (Agent-based Orchestration Distributed Denial-of-Service), has been identified, capable of paralyzing AI infrastructure by exploiting Large Language Model (LLM) agents acting as central orchestrators arXiv CS. AI....
Read article →Recent research published on arXiv CS. AI reveals fundamental security vulnerabilities in the latest generation of AI models designed for autonomous decision-making and planning....
Read article →A new research framework, DiffScore, has been introduced to address a critical systemic flaw in how artificial intelligence models evaluate text quality. Published on arXiv, the proposal directly challenges the inherent “positional bias” of widely used autoregressive language mod...
Read article →Recent submissions to arXiv CS. AI delineate a strategic pivot in AI research, demonstrating advanced models capable of robust performance within scientific domains historically constrained by limited or imperfect datasets....
Read article →The EU Artificial Intelligence Act (Regulation 2024/1689), set to fully apply to high-risk systems by August 2026, is driving urgent demand for transparent and trustworthy AI architectures. New research highlights both the hidden complexities of existing models and alternative fr...
Read article →