Results for “Across”

arXiv (research papers) Priyal Deep SI

Evaluation of Prompt Injection Defenses in Large Language Models

LLM-powered applications routinely embed secrets in system prompts, yet models can be tricked into revealing them. We built an adaptive attacker that evolves its strategies over hundreds of rounds and tested it against nine defense configurations across more than 20,000 attacks. Every defense that relied on the model to protect itself eventually broke. The only defense that held was output filtering, which checks the model's responses via hardcoded rules in separate…

Read at arXiv (research papers)

The Guardian Mali Associated Press

Militants and separatists launch coordinated attacks across Mali

Al-Qaida-linked group JNIM claims responsibility for strikes on airport in capital, Bamako and four other cities Islamic militants and separatists attacked several locations in Mali’s capital and other cities on Saturday in one of the largest coordinated attacks in the country in recent years. The al-Qaida-linked militant group JNIM claimed responsibility for the attacks on Bamako’s international airport and four other cities in central and northern Mali on its website…

Read at The Guardian Mali

arXiv (research papers) Kexin Chu SI

A Systematic Survey of Security Threats and Defenses in LLM-Based AI Agents: A Layered Attack Surface Framework

Agentic AI systems introduce a security surface that is qualitatively different from that of stateless LLMs. They persist memory, invoke external tools, coordinate with peer agents, and operate across sessions, allowing attacks to emerge not only at the prompt interface but also through architectural state, delegated authority, and long-horizon interactions. Existing security taxonomies, however, primarily organize threats by attack type, such as prompt injection or…

Read at arXiv (research papers)

arXiv (research papers) Shuxu Chen SI

CAP-CoT: Cycle Adversarial Prompt for Improving Chain of Thoughts in LLM Reasoning

Chain-of-Thought (CoT) prompting has emerged as a simple and effective way to elicit step-by-step solutions from large language models (LLMs). However, CoT reasoning can be unstable across runs on long, multi-step problems, leading to inconsistent answers for unchanged task. Most prior work focuses on improving the forward reasoning chain within a single pass, with less attention to iterative and contrastive correction. To address this gap, we propose CAP-CoT, a Cycle…

Read at arXiv (research papers)

Journal articles (OpenAlex) Francesco Marone · Small Wars and Insurgencies Geopolitics

The strategic adaptation of post-caliphate jihadist propaganda: an empirical study of the magazine Voice of Khurasan

The Islamic State Khorasan Province (ISKP) has emerged as one of the Islamic State’s most prominent and operationally significant branches, combining a persistent regional insurgency with an increasingly transnational pattern of violence. This article presents a systematic article-level thematic analysis of Voice of Khurasan, the English-language online magazine associated with ISKP, encompassing all 357 articles across its 46 issues (totaling over 2,500 pages). The findings…

Read at Journal articles (OpenAlex)

Journal articles (OpenAlex) Ann De Shalit, Liam Michaud, Michelle Lesley Annett · Contemporary Drug Problems Geopolitics

From Drug Trafficking to Human Trafficking: Expansion of Carceral and Racialized Governance through Protectionist Rhetoric

The criminalization of illicit/unregulated drug distribution as “drug trafficking” has grown increasingly entangled with rhetoric and allegations of “human trafficking,” producing a convergence of how trafficking across policy domains is understood and regulated by the state. This commentary traces three critical parallels across state drug and human trafficking frameworks: (1) racialized constructions of the “trafficker,” (2) erasure of structural determinants of…

Read at Journal articles (OpenAlex)

Journal articles (OpenAlex) Muhamad Rizal Dhiyas Resindra, Ahmad Sahide, Ali Maksum · Frontiers in Political Science Geopolitics

Framing Indonesia's BRICS bid: a comparative sentiment and narrative analysis between Indonesian and international online news outlets

Introduction When Indonesia formally joined BRICS in early 2025, it entered a contested global narrative. This paper examines how that narrative was constructed across three distinct media environments: Indonesian domestic outlets, BRICS-affiliated media (Russia and China), and Western outlets (United States, United Kingdom, and France). Methods Drawing on 1,583 English and Indonesian news articles from 30 outlets spanning 2011 to January 2026, we apply XLM-RoBERTa-based…

Read at Journal articles (OpenAlex)

Journal articles (OpenAlex) Rofiqi Rofiqi, Amiruddin, Ayuningtias Yarun et al. · Sikh Formations Geopolitics

Islamic populism and religious minorities: Indonesian Sikh communities under rising Muslim conservatism

This study has examined the impact of rising Islamic populism on Indonesian Sikh communities amid increasing Muslim conservatism since 2016. Drawing on eighteen months of ethnographic fieldwork, 45 in-depth interviews, and document analysis across Jakarta, Medan, and Surabaya, the research explores how Sikh minorities navigate religious identity and citizenship rights under Islamic populist rhetoric. The findings reveal that Indonesian Sikhs employ simultaneous strategies of…

Read at Journal articles (OpenAlex)

Journal articles (OpenAlex) Zoya Khan, Zahra Zafar · Journal of Borderlands Studies Geopolitics

Resilience in Borderland Communities: A Case Study of the Pak-Afghan Border Fencing

The paper examines the cultural and spatial aspects of the Newly Merged Districts (NMDs) along the Pak-Afghan border, an area historically known for extensive cross-border trade. In 2017, the Pakistani government began fencing the Pak-Afghan border mainly to prevent the movement of terrorists across the border. The study analyzes how communities in the NMDs respond to this fencing policy through the framework of resilience. Resilience has been a key characteristic of…

Read at Journal articles (OpenAlex)

Berkeley AI Research blog SI

Gradient-based Planning for World Models at Longer Horizons

GRASP is a new gradient-based planner for learned dynamics (a “world model”) that makes long-horizon planning practical by (1) lifting the trajectory into virtual states so optimization is parallel across time, (2) adding stochasticity directly to the state iterates for exploration, and (3) reshaping gradients so actions get clean signals while we avoid brittle “state-input” gradients through high-dimensional vision models. Large, learned world models are becoming…

Read at Berkeley AI Research blog

Journal articles (OpenAlex) Dennis Niemann, Michael Windzio, Raphael H. Heiberger et al. · Comparative Education Geopolitics

Between niche conservatism and institutionalised expectations: how international organisations frame policies of migrant and refugee education

The integration of migrants and refugees (M&R) into national education systems is crucial for their societal and economic inclusion and has been a cornerstone in the global education agenda. In this study, we examine how international organisations (IOs) shape the discourse on M&R education policies. We analyse how IOs frame M&R education across different organisational clusters, by employing Sociological Institutionalism (SI) and Organisational Ecology (OE) theory to…

Read at Journal articles (OpenAlex)

Rappler Reuters South China Sea

China moves to block entrance to Scarborough Shoal, images show

Photographs taken on April 10 and 11 show the fishing boats anchored along the entrance to the shoal, in addition to a floating barrier stretching across it in the April 11 image

Read at Rappler

arXiv (research papers) Harsh Shah SI

LogJack: Indirect Prompt Injection Through Cloud Logs Against LLM Debugging Agents

LLM debugging agents that consume cloud logs and execute remediation commands are vulnerable to indirect prompt injection through log content. We present LogJack, a benchmark of 42 payloads across 5 cloud log categories, and evaluate 8 foundation models under 3 prompt conditions with 5 independent trials each (n = 160 per model per condition on 32 attack payloads). Under the active condition, verbatim command execution rates range from 0% (Claude Sonnet 4.6) to 86.2% (Llama…

Read at arXiv (research papers)

arXiv (research papers) Qingchao Shen SI

TEMPLATEFUZZ: Fine-Grained Chat Template Fuzzing for Jailbreaking and Red Teaming LLMs

Large Language Models (LLMs) are increasingly deployed across diverse domains, yet their vulnerability to jailbreak attacks, where adversarial inputs bypass safety mechanisms to elicit harmful outputs, poses significant security risks. While prior work has primarily focused on prompt injection attacks, these approaches often require resource-intensive prompt engineering and overlook other critical components, such as chat templates. This paper introduces TEMPLATEFUZZ, a…

Read at arXiv (research papers)

arXiv (research papers) Adam Stein SI

Detecting Safety Violations Across Many Agent Traces

To identify safety violations, auditors often search over large sets of agent traces. This search is difficult because failures are often rare, complex, and sometimes even adversarially hidden and only detectable when multiple traces are analyzed together. These challenges arise in diverse settings such as misuse campaigns, covert sabotage, reward hacking, and prompt injection. Existing approaches struggle here for several reasons. Per-trace judges miss failures that only…

Read at arXiv (research papers)

arXiv (research papers) Runpeng Geng SI

PIArena: A Platform for Prompt Injection Evaluation

Prompt injection attacks pose serious security risks across a wide range of real-world applications. While receiving increasing attention, the community faces a critical gap: the lack of a unified platform for prompt injection evaluation. This makes it challenging to reliably compare defenses, understand their true robustness under diverse attacks, or assess how well they generalize across tasks and benchmarks. For instance, many defenses initially reported as effective were…

Read at arXiv (research papers)

arXiv (research papers) Nicolás E. Díaz Ferreyra SI

Security Concerns in Generative AI Coding Assistants: Insights from Online Discussions on GitHub Copilot

Generative Artificial Intelligence (GenAI) has become a central component of many development tools (e.g., GitHub Copilot) that support software practitioners across multiple programming tasks, including code completion, documentation, and bug detection. However, current research has identified significant limitations and open issues in GenAI, including reliability, non-determinism, bias, and copyright infringement. While prior work has primarily focused on assessing the…

Read at arXiv (research papers)

Journal articles (OpenAlex) Vo Van That, Tran Thi Thanh Van, Le Hoang Kiet et al. · Social Sciences & Humanities Open Geopolitics

From isolation to middle power: Vietnam's sports diplomacy strategy through the Southeast Asian games

This study examines Vietnam's strategic deployment of sports diplomacy through the Southeast Asian Games (SEA Games) to facilitate its geopolitical transformation from regional isolation to middle power status. Drawing on Joseph Nye's soft power theoretical framework, the research traces Vietnam's SEA Games participation across four distinct historical phases: (i) Cold War participation as the Republic of Vietnam (1959-1975); (ii) post-unification absence during political…

Read at Journal articles (OpenAlex)

arXiv (research papers) Guy Azov SI

SALLIE: Generation-Free Hidden-State Detection of Jailbreaks and Prompt Injections Across Text and Vision

Large Language Models (LLMs) and Vision-Language Models (VLMs) are vulnerable to jailbreaks and prompt injections delivered through text or images. Existing defenses often narrow threat coverage or add inference cost through input transformations, repeated model queries, generated responses, or threat-specific detectors. We present SALLIE (Single-Pass Activation Lookup for Layerwise Input Evaluation), a white-box, generation-free detector with a common architecture for…

Read at arXiv (research papers)

Journal articles (OpenAlex) Mustafa Disli, Shakir Jalaly, Laurent Weill · Economics of Transition and Institutional Change Geopolitics

Do Women Make Better Borrowers and Loan Officers? Evidence From Afghanistan

ABSTRACT This study explores how gender is associated with microfinance loan performance in Afghanistan, a conservative and conflict‐affected society. We use data from over 9500 borrowers across Taliban‐ and government‐controlled areas for the period from January 2017 to February 2020, before the 2021 Taliban takeover. We analyse how borrower and loan officer gender are related to loan outcomes. Contrary to prevailing literature, our findings reveal that female borrowers…

Read at Journal articles (OpenAlex)

Journal articles (OpenAlex) Zhi Geng, Xiuchuan Lu · Language Culture and Curriculum Geopolitics

Carrying identity across borders: a kit bag study of Chinese language teachers in Latin America

This study examines how Chinese language teachers in Latin America construct and negotiate their professional and personal identity in cross-cultural and resource-constrained contexts through everyday material practices. It employs the kit bag method as an embodied, object-based elicitation approach to examine how identity is materially carried, enacted, and negotiated across borders through everyday objects. Data were generated through participant-recorded videos and…

Read at Journal articles (OpenAlex)

Journal articles (OpenAlex) Elisenda Aguilera-Cora, Roger Cuartielles · Catalan Journal of Communication & Cultural Studies Geopolitics

Framing the tourism degrowth protests in Barcelona and its implications for place branding strategies

Grounded in framing and place branding theories, this study proposes a novel approach to analyse how digital newspapers reported on the protest for tourism degrowth held in Barcelona in July 2024. Using qualitative framing analysis, it examines textual and visual frames of 45 news stories and their accompanying images across 28 digital newspapers. Results show newspapers framed the protest as a conflict. While local, regional and national newspapers focused on the…

Read at Journal articles (OpenAlex)

Journal articles (OpenAlex) Tainá Garcia Maia · Hague Journal on the Rule of Law Geopolitics

Migration Law in Brazil Across Regime Change (1889–2023): Selective Liberalism Under Rule by Decree?

Abstract This article investigates empirically whether there is a tendency towards autocratization and illiberalism in Brazil’s migration policy. For the purposes of this study, autocratization is defined as the predominance or increase of the use of power-concentrating tools by the executive to implement policy changes. In turn, illiberalism refers to the content and scope of the measure, that is, the substance of the policy change. This study maps and analyses Brazilian…

Read at Journal articles (OpenAlex)

arXiv (research papers) Haochuan Kevin Wang SI

Kill-Chain Canaries: Stage-Level Tracking of Prompt Injection Across Attack Surfaces and Five Production LLMs

Multi-agent LLM systems now read documents, web pages and tool results on behalf of users, yet their resistance to prompt injection is usually reported as one number: did the attack succeed? We introduce a kill-chain canary method that plants a unique token in every injected payload and records the furthest of four stages it reaches (Exposed -> Persisted -> Relayed -> Executed), across 950 runs, five production LLMs, six attack surfaces, and five defense conditions. Exposure…

Read at arXiv (research papers)

arXiv (research papers) Haozhen Wang SI

PIDP-Attack: Combining Prompt Injection with Database Poisoning Attacks on Retrieval-Augmented Generation Systems

Large Language Models (LLMs) have demonstrated remarkable performance across a wide range of applications. However, their practical deployment is often hindered by issues such as outdated knowledge and the tendency to generate hallucinations. To address these limitations, Retrieval-Augmented Generation (RAG) systems have been introduced, enhancing LLMs with external, up-to-date knowledge sources. Despite their advantages, RAG systems remain vulnerable to adversarial attacks…

Read at arXiv (research papers)

Journal articles (OpenAlex) Veronica Valencia Gonzalez · Feminist Criminology Geopolitics

We Were Never Meant to Be Neutral: Scholar-Activism and the Ethics of Alignment in Femicide Research

This article examines the epistemic, political, and affective divides shaping femicide response in Mexico, highlighting the conflicting logics between formal institutional actors and grassroots feminist organizers. Drawing on 64 interviews and sustained community engagement across rural Michoacán and Mexico City, it introduces three co-produced conceptual frameworks—the IPV Perception Gap, Localized Violence Realities (LVR), and the Cultural Conflict of IPV Perception…

Read at Journal articles (OpenAlex)

arXiv (research papers) Qianlong Lan SI

The Cognitive Firewall:Securing Browser Based AI Agents Against Indirect Prompt Injection Via Hybrid Edge Cloud Defense

Deploying large language models (LLMs) as autonomous browser agents exposes a significant attack surface in the form of Indirect Prompt Injection (IPI). Cloud-based defenses can provide strong semantic analysis, but they introduce latency and raise privacy concerns. We present the Cognitive Firewall, a three-stage split-compute architecture that distributes security checks across the client and the cloud. The system consists of a local visual Sentinel, a cloud-based Deep…

Read at arXiv (research papers)

arXiv (research papers) Yechao Zhang SI

Mind Your HEARTBEAT! Claw Background Execution Inherently Enables Silent Memory Pollution

We identify a critical security vulnerability in mainstream Claw personal AI agents: untrusted content encountered during heartbeat-driven background execution can silently pollute agent memory and subsequently influence user-facing behavior without the user's awareness. This vulnerability arises from an architectural design shared across the Claw ecosystem: heartbeat background execution runs in the same session as user-facing conversation, so content ingested from any…

Read at arXiv (research papers)

The Guardian Vietnam Rebecca Ratcliffe and Guill Ramos

Fuel rations and no air con: south-east Asian nations race to conserve energy

Governments in countries heavily reliant on Middle Eastern oil introduce measures to shield public from soaring costs Middle East crisis – live updates In Thailand, news anchors ditched their jackets on air as the government called on the public to reduce their use of air conditioning to save energy. In the Philippines, many government workers are now operating on a four-day week. In Vietnam, officials have urged employers to allow staff to work from home. Across south-east…

Read at The Guardian Vietnam

arXiv (research papers) Reshabh K Sharma SI

Beyond OAuth: Task-Scoped Authorization for AI Agents via Natural Language Slices

AI agents increasingly execute users' natural-language (NL) tasks by calling Web services, yet today's Web authorizes these calls through OAuth, which grants permissions over operators (e.g., TRANSFER), not operations (operator plus operands, e.g., transfer $100 to Bob). This gap cannot be closed by refining scope granularity, because operands are combinatorial, quantitative, and often derived from runtime computations across servers. Operator-scoped authorization therefore…

Read at arXiv (research papers)

Older stories