Importance. Ensuring flight safety is a priority for air transport. Human error, particularly communication failures in the pilot-controller system, is a critical risk factor. Even with the use of strictly regulated ICAO phraseology, verbal interactions are subject to deviations caused by stress, cognitive load, or language barriers, necessitating the development of new methods for analyzing and monitoring aviation discourse. Research Methods. The study is based on a comprehensive linguistic approach that examines aviation communication at the lexical (deviations from standards), prosodic (tone, tempo, and pause analysis), and pragmatic (speech act analysis) levels. This approach allows for the development of a theoretical and methodological foundation for the application of machine learning and computational linguistics. Results and Discussion. Drawing on recommendations of scientific and methodological literature, this paper analyzes the potential of modern machine learning (ML) methods and models to address human-factor-related flight safety issues. In particular, the paper examines the specifics of using automatic speech recognition (ASR), topic modeling (LDA, BERTopic), and classification architectures in this subject area: transformer (BERT), hybrid, and classical ML models based on embeddings. It is demonstrated that modern algorithms are capable of detecting not only overt protocol violations but also implicit stress markers (changes in pitch) and pragmatic mismatches (discrepancies between intention and perception). A comparative specifics analysis of using machine learning classification models to address aviation safety issues is conducted. A retrospective example of the Avianca Flight 052 disaster is used to demonstrate how multimodal ML analysis could have proactively identified a developing critical situation based on linguistic and acoustic anomalies. The article describes the development prospects for this area, related to the creation of integrated intelligent systems. Conclusions. The symbiosis of linguistic ontology and the modern capabilities of machine learning methods creates a new paradigm for proactive aviation safety. This enables a transition from post-factum incident analysis to the creation of intelligent decision support systems for air traffic controllers, objective assessment of pilot linguistic proficiency, and the identification of latent risks in large text datasets.

Toward a relational paradigm for AI safety: relational synchrony in the heart-inspired dual-layer architecture (HIDLA)

Orthokeratology (OK) lenses for myopia correction are susceptible to biofouling by tear-derived biomolecules, escalating the risks of ocular infection and inflammation. Prevailing studies that rely on end point protein quantification fail to capture the real-time kinetics of fouling formation. Here, we fabricated a UV pressure-assisted polymer-grafted quartz crystal microbalance with dissipation monitoring (QCM-D) sensor exhibiting exceptional stability, nanoscale smoothness (RMS roughness ≈2 nm), and interfacial peel resistance. This platform enables in situ tracking of adsorption/desorption kinetics for four critical tear components: native/denatured lysozyme and oxidized/native lecithin. Key findings reveal a flow-dependent fouling behavior, wherein low flow rates increase biomolecule adsorption by 37-80% compared with higher flows. We further identify denatured lysozyme and native lecithin as resilient contaminants characterized by a stronger deposition affinity and pronounced resistance to elution. Quantitative screening of multipurpose solutions (MPSs) demonstrates that MPS #2 achieves 20-100% elution rate across biomolecules, outperforming commercial benchmarks. By leveraging dissipation-frequency (D-F) analysis, we clarify the fundamental mechanisms of biofouling formation at the molecular level. Collectively, this work establishes three critical advances: (1) a real-time biofouling diagnostic platform for OK lens interfaces, (2) molecular design principles for antifouling materials based on adhesion remodeling theory, and (3) an accelerated MPS formulation screening paradigm for ocular device safety.

To meet the long-term safety demands of gravity-based platforms in marine ranching, this paper develops a structural health monitoring system based on digital-twin model pattern matching (DT-PM-SHM@GRP). The system is built around a four-closed-loop architecture of “sensing–twinning–matching–decision”. First, an integrated sensor network combining fiber Bragg gratings, inclinometers and meteorological instruments is deployed on site to establish a multi-source synchronous acquisition and processing system. Second, a 3D adaptive FE twin model is created; by coupling it with all potential anticipated environmental loads, a database set that links environmental parameters–load cases–structural responses is generated. Real-time on-site data features extracted from the sensor network are then used to search the twin model states via pattern matching, enabling reliable tracking of the twin’s condition. On this basis, online assessment of structural deformation, stress distribution and overall platform integrity assessment is carried out, and predictive decision-making suggestions are provided. The developed system offers a transferable and scalable digital-twin monitoring paradigm for intelligent marine-ranch safety and maintenance.

Memory characteristics in large language models (LLMs) represent a transformative progress that enables relevant continuity, privatization, and adaptive learning in interactions. However, these capabilities introduce novel security vulnerabilities that extend beyond traditional concerns. This article examines the security implications of memory-enabled LLMs, categorizing architectural approaches and identifying distinct vulnerability classes, including temporal prompt injection, information persistence, and memory poisoning. Through documented case studies and empirical evidence, the article illustrates how these vulnerabilities manifest in production environments, leading to data leakage, system manipulation, and knowledge corruption. The article proposes comprehensive security frameworks incorporating memory segregation, temporal constraints, bidirectional filtering, differential privacy, and advanced auditing mechanisms. Since LLMS develops from stateless tools to constant assistants, safety paradigms must expand the traditional boundaries to address the entire memory lifestyle and ensure that these systems remain both functional and safe in sensitive operating contexts.

Rapid urban air mobility (UAM) developments and new classes of vertical takeoff and landing (eVTOL) aircraft have changed the safety paradigm in urban airspace. eVTOL aircraft operations in dense urban environments are characterized by increased variability of external factors, highly dynamic flight scenarios, and an increased likelihood of rare but potentially critical events. Traditional safety assessment approaches do not capture the specific features of eVTOL designs, power plants, autonomy algorithms, and urban air traffic characteristics; this results in low threat prediction accuracy and limited development of modern incident prevention systems. Herein, the risk profile of eVTOL aircraft is analyzed, accounting for the multifactorial nature of urban environments and the complexity of integrating such vehicles into existing UAM infrastructure. The need for quantitative methods for assessing the probability of critical situation risks is also substantiated. These methods provide a statistically accurate description of extreme events and enable the identification of hidden dependencies in complex technical and organizational systems. Approaches based on probabilistic models, extreme value analysis, and systemic

Sulfide all solid-state batteries represent a promising next generation energy storage technology. However, their presumed safety is challenged by the risk of thermal runaway initiating at unexpectedly low temperatures. This critical issue stems from the unstable chemical interface between the positive electrode and thiophosphate solid electrolyte, a factor often overlooked in favor of electrochemical studies. Here we demonstrate that this electrochemically formed interphase is the primary trigger for catastrophic failure, not the bulk materials. Our investigation reveals a universal two stage degradation mechanism. The first stage involves intense exothermic reactions at the interface below 160 °C, releasing heat and gases. This initiates a second stage of propagating reactions leading to thermal runaway. Crucially, we show this hazardous process can be suppressed by interface engineering. We design a stable interfacial layer using a germanium sulfur chemistry, specifically lithium germanium sulfide. This modification delivers improved thermal safety without sacrificing battery performance. Our findings have the potential to establish a forward-looking safety paradigm, shifting the focus from bulk material compatibility to interfacial stability, and provide a vital design principle for future safe solid-state batteries.

Ensuring safety in robotic manipulation is increasingly critical as robots become integrated into human-shared environments for complex physical interaction tasks. This paper presents an energy-aware control framework that combines active responses with passive compliance for safety-critical robotic manipulation. Specifically, Control Barrier Functions (CBFs) are employed for active collision avoidance with detected obstacles, which are then integrated with fallback safety actions to resolve potential violation of CBF constraints. Complementing this active safety paradigm, a passive safety paradigm is implemented to mitigate post-collision impacts by monitoring energy variance and limiting power exchanges. Furthermore, an energy tank is incorporated to enforce passivity of the robot, which is crucial to address potential instability issues in variable impedance control. To make the tank adaptive to varying energy requirements arising from dynamic environments and unpredictable events, we propose a novel, task-agnostic tank recharging condition without compromising the system's passivity guarantee. The effectiveness of the proposed control framework is validated through experiments on a KUKA iiwa 14 robot.

Ensuring robust railway safety is paramount for efficient and reliable transportation systems, a challenge increasingly addressed through advancements in artificial intelligence (AI). This review paper comprehensively explores the burgeoning role of AI in enhancing the safety of railway operations, focusing on key contributions from machine learning, neural networks, and computer vision. We synthesize current research that leverages these sophisticated AI methodologies to mitigate risks associated with railroad accidents and optimize railroad tracks management. The scope of this review encompasses diverse applications, including real-time monitoring of track conditions, predictive maintenance for infrastructure components, automated defect detection, and intelligent systems for obstacle and intrusion detection. Furthermore, it delves into the use of AI in assessing human factors, improving signaling systems, and analyzing accident/incident reports for proactive risk management. By examining the integration of advanced analytical techniques into various facets of railway operations, this paper highlights how AI is transforming traditional safety paradigms, paving the way for more resilient, efficient, and secure railway networks worldwide.

Bullying in educational settings remains a critical issue, particularly for students with Autism Spectrum Disorder (ASD) who often struggle with verbal communication and social cues. This research addresses the lack of affordable and non-invasive early warning systems designed specifically for the sensory needs of autistic children. The primary objective was to develop and test an Arduino- and Internet of Things (IoT)-based early warning bell system to empower ASD students to report bullying incidents in real time. Using a Research and Development (R&D) approach with the 4D development model (Define, Design, Develop), the system was tested by 50 stakeholders, including special education teachers and experts. Results indicate a high level of system effectiveness, with a 94% rapid warning response rate and a 90% increase in students' perceived sense of safety. Statistical analysis yielded a Cronbach’s Alpha of 0.948, confirming high reliability. This study concludes that integrating simple IoT push-button technology with Telegram notifications provides a feasible, low-cost solution for inclusive schools. It shifts the safety paradigm from passive monitoring to active student empowerment, offering a sustainable protection strategy for vulnerable learners.

The present work offers a comprehensive, systematic study of the evolution of nail service technologies over the period from 2011 to 2024. The study is based on the integration of global market data, dermatological research on the pathogenesis of nail apparatus injuries, and an empirical 13-year experience in implementing the proprietary wet grinding technique in clinical and aesthetic practice. The central hypothesis of the study is that the adaptation of podological spray technologies into aesthetic manicure protocols mitigates the risks of thermal necrosis of the matrix and mechanical traumatization of the eponychium inherent in popular techniques of dry hardware manicure. Particular attention is paid to the analysis of the safety of the method for patients with diabetes mellitus and impaired tissue trophics. Using the case study of the Nailslab salon, the economic efficiency (reduction of procedure time to 1 hour 40 minutes) and clinical safety of the proposed protocol are demonstrated. The study confirms the need to revise educational standards in the industry with an emphasis on hydro-mechanical treatment methods.

Rapid urbanization and intensifying climate volatility have subjected United States municipalities to complex, compound risks, ranging from catastrophic flooding and wildfires to heat-induced infrastructure failures. Traditional risk assessment models often rely on static physical data, frequently failing to account for the dynamic spatiotemporal nature of these hazards and the socioeconomic disparities that define community vulnerability. This paper proposes a novel Cross-Domain GeoAI Framework, termed GeoRes-X, which integrates multi-source Geographic Information Systems (GIS), remote sensing environmental data, and social vulnerability metrics into a unified deep learning architecture. The objective of this study is to enhance multi-hazard resilience modeling and facilitate equitable emergency governance. By employing a Spatiotemporal Graph Neural Network (ST-GNN), the framework unifies high-resolution satellite imagery analysis with complex infrastructure network reasoning to predict failure points and prioritize resource allocation. The methodology utilizes real-world datasets comprising infrastructure topology, historical hazard events, and demographic census data. The results demonstrate that GeoRes-X significantly outperforms traditional logistical regression and static machine learning benchmarks in predicting infrastructure stress, achieving a prediction accuracy of 92.4%. Furthermore, the inclusion of equity constraints reduces algorithmic bias against marginalized communities by 18%, ensuring that emergency management decisions prioritize high-risk, low-resource areas. This research bridges the gap between technical hazard mapping and ethical decision-making, offering a transformative paradigm for public safety.

The Rust programming language has garnered significant attention due to its robust safety features and memory management capabilities. Despite its guaranteed memory safety, Rust programs suffer from runtime errors that are unmanageable, i.e., panic errors. Notably, traditional memory issues such as null pointer dereferences, which are prevalent in other languages, are less likely to be triggered in Rust due to its strict ownership rules. However, the unique nature of Rust's panic bugs, which arise from the language's stringent safety and ownership paradigms, presents a distinct challenge. Over half of the bugs in rustc, Rust's own compiler, are attributable to crashes stemming from panic errors. However, addressing Rust panic bugs is challenging and requires significant effort, as existing fix patterns are not directly applicable due to the design and feature of Rust language. Therefore, developing foundational infrastructure, including datasets, fixing patterns, and automated repair tools, is both critical and urgent. This paper introduces a comprehensive infrastructure, namely PanicFI , aimed at providing support for understanding Rust panic bugs and developing automated techniques. In PanicFI , we construct a dataset, Panic4R , comprising 102 real panic bugs and their fixes from the top 500 most downloaded open-source crates. Then, through an analysis of the Rust compiler implementation, we identify Rust-specific patterns for fixing panic bugs, providing insights and guidance for generating patches. Moreover, based on these patterns, we develop an automated fixing tool, namely PanicKiller , as an artifact, which has already contributed to the resolution of 28 panic bugs in open-source projects. The practicality and efficiency of PanicKiller confirm the effectiveness of the patterns mined within PanicFI . Furthermore, Panic4R serves as a benchmark for evaluating APR tools focused on Rust panic bugs. We believe the construction and release of PanicFI could enable the expansion of automated repair research tailored specifically to Rust programs, addressing unique challenges and contributing significantly to advancements in this field.

Resilience represents a crucial safety paradigm characterizing system risk evolution under specific perturbations, yet its application to microscopic driving safety needs further exploration. Analogous to resilience evolution patterns, generalized driving risk progression exhibits distinct safety decay and recovery phases, thereby establishing theoretical foundations for driving safety resilience analysis. This study develops a resilience-centric framework for quantifying dynamic car-following risk evolution. The car-following pairs were selected from highD natural driving trajectory dataset. Following risk quantification and clustering, 1,676 samples demonstrating complete “safe-dangerous-safe” transition process were identified. Through decomposition of car-following resilience into safety decay and recovery phases, we derived five interpretable resilience features and modeled their functional dependencies on driving variables via an integrated machine learning and SHapley Additive exPlanations (SHAP) analytical framework. Key findings reveal: 1) the CatBoost demonstrated superior fitting performance, achieving a mean absolute percentage error less than 10% across all resilience features; 2) the proposed Low-Rank Polynomial SHAP Fitting (LRP-SF) captured the nonlinear relationships between driving variables and resilience features, quantifying both directional influences and threshold effects; and 3) driver risk perception exhibits phase-dependent variability, with a stimulus-response mechanism governing safety evolution dynamics. The rapid safety deterioration and hazardous states act as triggers for recovery processes. This study further examined resilience threshold validity, car-following variable interactions, and inter-feature correlations. The potential application of LRP-SF in car-following safety control was also anticipated. This study offers methodological advancements for advanced driver assistance systems development and establishes a novel analytical paradigm for driving safety research.

Norfloxacin (NFX) residues in food and environmental samples threaten human health, necessitating rapid on-site detection. A hierarchical MOF@MOF probe (Ag-MOF-74/NH₂-MIL-53(Al) for ratiometric NFX sensing has been fabricated, in which NH₂-MIL-53(Al) enhances NFX emission via hydrogen bonding and energy transfer, while Ag-MOF-74 provides a stable reference, enabling self-calibration. The probe can not only achieve ultrasensitive detection within 75s, but also exhibits a low limit of detection (LOD), high selectivity and good reproducibility. A smartphone-based platform was designed for on-site quantification within minutes. Validated in milk, soil, and water, it showed excellent accuracy (98-102% recovery) and precision (RSD < 2.86%). This MOF@MOF strategy offers a robust on-site screening tool and a versatile design paradigm for food safety and environmental monitoring.

In the quarter of a century since the release of the report To Err is Human, current progress in patient safety is at best inconsistent, and at worst, has outright stalled. To resume the speed of progress made at the start of the patient safety movement, we will need to approach patient safety in a different way. The lack of progress in patient safety has increased enthusiasm for different paradigms of understanding patient safety. Rather than focusing on deficit-based models of patient safety, newer approaches focus on complementary methods that attempt to understand the essential underpinnings of work that is safe. Weick describes the story of wildland firefighters who failed to drop their tools when they were no longer useful, resulting in their deaths. While the tools used by patient safety professionals are not physical implements, a similar phenomenon exists. In this review, the commonly used patient safety tools which are impeding progress are discussed. Alternative views to citing human error as a cause, ruthlessly targeting unachievable goals, and approaching the complex environment of healthcare as a linear system are presented.

The construction sector continues to be one of the world's most hazardous, with high rates of accidents fuelled by multicomponent site dynamics, extensive use of heavy equipment, and unstable human behaviour. Conventional safety management methods, although essential, are generally reactive and fall short in offering real-time hazard perception or forecasting risk assessment. Recent advancements in Artificial Intelligence (AI) provide revolutionary opportunities to enhance safety performance through anticipatory, automated, and evidence-based decision-making. This article explains how AI techniques—ranging from computer vision for PPE detection and unsafe behaviour recognition, to wearable sensor analysis for fatigue and stress monitoring, to predictive machine learning models for incident prediction—can significantly enhance construction safety management. Furthermore, the combination of AI with Building Information Modelling (BIM) and digital twin technology allows for real-time hazard mapping, safety scenarios through simulation, and end-to-end synchronization between the virtual and physical worlds. This paper proposes a complete AI-based safety paradigm that harmonizes multimodal data sources, edge analytics, and interpretable predictive models to close the risk mitigation gap with worker privacy and trust. Data quality anomalies, model generalization, alert fatigue, and surveillance implications in terms of ethics are also addressed with responsible deployment practices. AI will eventually be able to shift construction safety from reactive compliance to preventive intervention, reducing incidents and safer conditions.

INTRODUCTION: Surgical intervention is a cornerstone in the management of moderate to advanced glaucoma, a leading cause of irreversible blindness worldwide. The surgical landscape has evolved from traditional incisional procedures, such as trabeculectomy and glaucoma drainage device (GDD) implantation, to a diverse array of minimally invasive glaucoma surgery (MIGS) techniques. This evolution reflects a continuous effort to balance the efficacy of intraocular pressure (IOP) reduction with the risk of surgical complications. This systematic review aims to comprehensively synthesize and compare the complication profiles associated with this full spectrum of modern glaucoma surgeries. METHODS: A systematic literature search was conducted in PubMed, Google Scholar, Semantic Scholar, Springer, Wiley Online Library for studies published between January 2000 and October 2024. The search included randomized controlled trials (RCTs), prospective and retrospective cohort studies, and large case series reporting on complications of trabeculectomy, GDDs, and various MIGS procedures. Two independent reviewers performed study selection and data extraction. The methodological quality of RCTs was assessed using the Cochrane Risk of Bias 2 (RoB 2) tool. Data on over 15 distinct intraoperative, early postoperative, and late postoperative complications were extracted and narratively synthesized. RESULTS: A total of 22 studies, including 8 major RCTs and 14 cohort studies, met the inclusion criteria. Traditional surgeries were associated with higher rates of severe complications. The 5-year results of the Primary Tube Versus Trabeculectomy (PTVT) Study showed that early postoperative complications were significantly more frequent after trabeculectomy than tube shunts (34% vs. 19%, p=0.013), though rates of late and serious complications were similar by year five. Comparative trials of GDDs, such as the Ahmed Baerveldt Comparison (ABC) and Ahmed Versus Baerveldt (AVB) studies, demonstrated that non-valved Baerveldt implants had a higher incidence of hypotony-related complications compared to valved Ahmed implants, which were more prone to hypertensive phase and encapsulation. MIGS procedures demonstrated a markedly more favorable safety profile. The 5-year HORIZON trial found the cumulative risk of subsequent incisional surgery was significantly lower with the Hydrus microstent plus phacoemulsification compared to phacoemulsification alone (2.4% vs. 6.2%, p=0.027), with no long-term adverse safety signals. The most common complications for MIGS were transient hyphema and IOP spikes. DISCUSSION: The evidence confirms a distinct trade-off between surgical efficacy and safety. Traditional procedures offer the most substantial IOP reduction but carry a significant risk of vision-threatening complications like bleb-related endophthalmitis and refractory hypotony. MIGS procedures offer a safer alternative, particularly for mild-to-moderate glaucoma, primarily reducing medication burden with minimal risk of severe adverse events. The choice of surgery is therefore dependent on a nuanced assessment of the patient's disease severity, target IOP, and risk tolerance. The inconsistent reporting of complications across studies remains a significant barrier to direct meta-analytic comparison. CONCLUSION: Glaucoma surgery encompasses a spectrum of procedures with widely varying complication profiles. While traditional surgeries remain indispensable for advanced disease, MIGS has fundamentally improved the safety paradigm for patients with less advanced glaucoma. Future research must adopt standardized definitions and reporting protocols for complications to allow for more robust evidence synthesis and to better guide clinical decision-making.

Antimicrobial resistance and climate change in the One Health food safety paradigm: A global perspective.

Lithium metal batteries (LMBs) are expected to increase energy density due to the high capacity and low electrode potential of lithium metal. However, lithium dendrite growth and organic liquid electrolytes exacerbate the risk of thermal runaway. To improve the safety of the battery, a multifunctional flame-retardant separator was developed through the synergistic effect of decabromodiphenyl ethane (DBDPE)/Al2O3 nanoparticle composite modification. Among these, DBDPE acts as a flame retardant to reduce the combustibility of the battery, while Al2O3 with high mechanical strength inhibits dendrite growth, and its amphiphilic nature favors the uniform distribution of lithium ions. Thus, a multifunctional flame-retardant separator simultaneously achieves excellent flame suppression, enhanced thermal conductivity (71.3 mW·m-1 k-1), and excellent electrochemical performance. Lithium (Li) symmetrical batteries based on this separator stably run over 500 h at a current density of 0.5 mA·cm-2, and Li/LiFePO4 batteries retain a capacity of 138 mAh g-1 capacity over 100 cycles at a rate of 0.5 C. This flame-retardant separator design, leveraging synergistic regulation of microstructure and thermal properties, provides a groundbreaking roadmap for suppressing thermal runaway while maintaining electrochemical performance, thereby redefining safety paradigms for next-generation high-energy-density battery systems.