共找到 20 条结果
The last months of 2025 have brought major new policy initiatives and pilots for digital medicine and AI regulatory sandboxes from the EU, UK, and US regulators. The EU has seen a regulatory simplification package with the promotion of sandboxes for innovative technologies. The UK sees the start of an important second phase of the world-leading ‘AI Airlock’ regulatory sandbox program. TEMPO, a major voluntary alternative pathway regulatory innovation pilot for digital health in chronic disease has been announced by the US FDA. These frameworks bring flexible approaches to explore digital health technology innovation together with regulatory innovation. They will all substantially advance digital medicine, with the US TEMPO project being the most innovative, bringing sandboxing principles to the on-market phase.
African governments must regulate fast-growing digital health and artificial intelligence technologies while building the continent's planned digital single market. Regulatory sandboxes are increasingly promoted as a solution, but most sandbox experience comes from finance, where the primary concern is market risk, not clinical safety or public health harm. Healthtech is not fintech, and failure carries very different consequences. We argue that the Global System for Mobile Communications (GSM) standards model offers a more appropriate template. Drawing on the V-Model framework, we propose that regional bodies verify technical interoperability while national regulators validate clinical and ethical suitability. Simulation-based exercises can test governance before technologies go live, building a cautious, collectively owned approach to digital health innovation on African terms.
Recent technological advances in artificial intelligence and machine learning have led to rapid changes in clinical informatics. With the release of new models, the barrier to creating medical software has markedly decreased. As noted in the now-viral article by Matt Shumer, we have moved from a world where building a clinical application took a team of engineers working over the course of months to years, to one where a single clinician can develop and deploy a working prototype in hours to days. However, the same innovation driving this 'explosion' of tools also carries the risk of flooding health systems with unvalidated 'AI slop.' As prohibiting tools developed through AI runs the risk of stifling innovation and limiting potentially impactful discoveries, we suggest abandoning the slow, fear-based resistance to AI adoption. Further, we propose a 'go fast and fix things' framework-utilising target product profiles, regulatory sandboxes, and continuous audit-to harness the speed of generation and prototyping while rigorously validating clinical utility.
The secondary use of electronic health records (EHRs) poses legal challenges, particularly when the responsibility for managing EHRs lies with local or regional authorities. This article presents a case-based analysis of the secondary use of EHR data in contexts where data privacy responsibilities are managed regionally in Sweden. Using two distinct purposes for the secondary use of the digital tool Patient Overview Breast Cancer: (i) assessing the uptake of new treatment strategies in a real-world setting for quality assurance, and (ii) evaluating the effectiveness of these strategies in specific patient subgroups with limited evidence for research purposes, the study explored the distinctions between research and quality assurance, the legal implications of each framework, and the potential role of federated learning as a privacy-preserving technological solution. Federated learning offers a promising approach to overcome legal and organizational barriers to secondary use of regional EHRs in Sweden, enabling scalable, clinically meaningful insights for cancer care. However, its effective implementation requires a unified national framework that balances personal integrity with patient safety, supported by regulatory sandboxes.
Artificial intelligence (AI) transforms extreme-weather forecasting by delivering faster and more accurate predictions at a fraction of the computational cost of traditional models. However, these advances are often accompanied by opaque decision processes, raising challenges for trust, equity, and long-term resilience in early warning systems. This article examines transparency in AI-based forecasting across three dimensions-predictive integrity, societal fairness, and long-term resilience-and argues that accuracy alone is insufficient in high-stakes contexts. Drawing on recent regulatory developments and global meteorological practice, we outline practical measures such as harmonized forecast labeling, impact-ready model cards, and extreme-event regulatory sandboxes. Embedding these measures within international frameworks is essential to ensure that the speed and efficiency of AI-driven forecasts translate into effective, trusted, and equitable early warning systems.
Domestic cats (Felis catus) are ubiquitous companion animals that provide substantial psychological and social benefits to children and adults alike, but they also serve as reservoirs and vectors for a wide range of zoonotic pathogens. Close physical contact between cats and children, frequent use of shared environments such as homes, playgrounds, and sandboxes, and still-developing hygiene behaviours increase opportunities for exposure to protozoa, helminths, bacteria, fungi, and ectoparasite-borne agents. This review synthesizes current evidence on key feline-associated zoonoses of pediatric concern-including Toxoplasma gondii, Toxocara cati, Ancylostoma spp., Dipylidium caninum, Bartonella henselae, Salmonella enterica, Campylobacter jejuni, Pasteurella multocida, Microsporum canis, flea-borne Rickettsia species, and rabies-with emphasis on transmission routes, clinical manifestations, and risk modifiers in children, pregnant women, and immunocompromised individuals. Within a One Health framework, we also summarize global publication trends on feline zoonoses, discuss how urban cat ecology and management (including free-ranging cats in child-frequented environments) may shape pediatric risk, and outline practical prevention strategies centred on hygiene, veterinary care, and targeted education for caregivers and children.
The increasing frequency of non-stationary hydrological extremes is outpacing the response capacity of conventional rule-based watershed management, leading to elevated risks of algal blooms and contaminant pulses in river-reservoir systems. Addressing these dynamic environmental risks necessitates a transition toward adaptive, closed-loop decision-making. Beyond ongoing challenges in data scarcity and model transferability, a critical operational bottleneck is the gap between reliable prediction and safe physical intervention. To address the translation of model outputs into physical control actions without violating safety boundaries, this review establishes a maturity framework linking artificial intelligence capabilities to the safe delegation of decision authority. Three core contributions are presented. First, physics-informed machine learning and spatiotemporal graph networks are synthesized to show how physical conservation laws overcome data scarcity and topological complexity. Second, the closed-loop decision layer is critically examined through targeted operational scenarios: virtual sandboxes for preemptive algal bloom mitigation, multi-agent reinforcement learning for dynamic flood coordination, and inverse modeling for rapid chemical spill containment. Third, deployment barriers are distilled into four foundational pillars: computational infrastructure, algorithmic robustness, cyber-physical security, and institutional accountability. Finally, an open benchmark ecosystem is proposed to operationalize this shift, advancing watershed digital twins from passive monitors into responsible intelligent agents.
Ontario faces persistent diagnostic imaging (DI) challenges, which includes fragmented implementation of Artificial Intelligence (AI) solutions. The Canadian Association of Radiologists (CAR) has emphasized the need for freely available, modality-specific sandboxes to safely evaluate and refine AI tools prior to clinical use. In response, this paper proposes a provincially governed AI Sandbox for Diagnostic Imaging, designed to enable secure testing, validation, and scaling of AI innovations within a privacy-by-design framework. The AI Sandbox integrates six policy domains, such as clinical utility, ethics and privacy, technical design, governance, legal compliance, and sustainability, to guide responsible adoption. Supported by the Ontario Ministry of Health's commitment to trustworthy AI, this initiative aims to foster equitable, transparent, and system-wide innovation in diagnostic imaging.
Digital assets (DAs) such as cryptocurrencies, tokenized securities, stablecoins, non-fungible tokens (NFTs), and central bank digital currencies, are transforming financial markets with new business models, investment opportunities, and transaction efficiencies. Underpinned by blockchain, distributed ledger technology, and smart contracts, digital innovations are reshaping the financial ecosystem. However, their rapid growth introduces substantial risks, including fraud, market manipulation, cybersecurity threats, and regulatory uncertainty. This position paper offers an interdisciplinary and empirically grounded analysis of the DA landscape. We define and classify major asset types, trace their evolution from speculative instruments to functional tools, and assess current adoption trends. Additional technological developments (e.g., decentralized finance and NFT expansion) are examined for their role in accelerating this transformation. We also analyze the global regulatory landscape, highlighting jurisdictional differences, classification challenges, and emerging governance frameworks. To address key risks, we derive mitigation strategies via quantitative analysis and case-based evidence. The risks include balancing innovation with investor protection through adaptive regulatory design, promoting cross-border regulatory harmonization to prevent arbitrage and fragmentation, and supporting experimentation through regulatory sandboxes and innovation hubs. By adopting a forward-looking, evidence-based, and collaborative regulatory approaches, stakeholders can harness the benefits of DAs while managing systemic risks and maintaining market integrity.
Analytical sciences increasingly require autonomous systems capable of real-time heuristic reasoning and dynamic parameter adaptation. However, current automated laboratory platforms remain largely restricted to rigid synthesis workflows. Here, we introduce the Artificial Raman Expert (ARE), an autonomous AI-driven analytical framework powered by localized, open-source large language models (LLMs) to ensure strict data privacy and reproducibility. By integrating a structured knowledge base (A-Knowhow), ARE internalizes the tacit decision-making logic of seasoned spectroscopists. To ensure safe and scalable deployment, ARE's cognitive and optical reasoning was first validated within a risk-free virtual reality (VR) sandbox before being successfully translated to a physical Raman spectrometer. Across highly complex physicochemical scenarios, ARE demonstrated advanced cognitive capabilities: evaluating the spatial heterogeneity of pharmaceutical samples, performing spectral unmixing to isolate trace narcotics from severe forensic matrix interferences within just 3 analytical cycles, heuristically halting analyses after 8-10 iterations-despite human prompts mandating >30 positions-thereby reducing analytical time and resource consumption by ∼70%. Additionally, ARE conducted constraint-aware parameter tuning, dynamically capping laser exposure at 60 s to optimize signal-to-noise ratios while strictly preventing biological cell phototoxicity. By actively optimizing analytical resources and adapting to noisy data, ARE demonstrates the feasibility of context-aware autonomous spectroscopy. This work provides a foundational proof-of-concept for encoding tacit analytical expertise into AI-driven instrumentation, with potential for broader application across diverse analytical domains.
Very High Frequency (VHF) radio communication systems face significant challenges in modern electromagnetic environments, including spectrum congestion, dynamic interference, and varying channel conditions. Existing adaptive approaches rely on static rule-based switching or single-cycle optimization, which cannot accumulate operational experience across decision cycles. This paper proposes a digital twin-enabled online learning framework (DT-MAB) for adaptive waveform selection in tactical VHF communication. The framework employs a contextual multi-armed bandit algorithm (Lin-UCB) that continuously learns the mapping from channel conditions to optimal configurations, with the digital twin serving as a virtual exploration sandbox that screens candidate configurations before physical deployment-preventing link disruptions during exploratory actions. An expanded configuration space of 63 candidates (7 waveforms × 3 MAC protocols × 3 power levels) is constructed, and a hierarchical performance evaluation model combining voice quality, bit error rate, communication delay, and transmission range is developed using the Analytic Hierarchy Process (AHP) as the reward function for online learning. Experimental results across 10 random seeds demonstrate that DT-MAB achieves the lowest mean cumulative regret, reducing regret by 29% relative to MAB without a digital twin and by 16.5% relative to PSO-based optimization on average. Ablation experiments confirm that removing virtual exploration increases performance drop events by 49% (from 250 ± 79 to 373 ± 6), demonstrating that the digital twin is a functionally indispensable component of the online learning architecture.
The rapid growth of Internet-of-Things (IoT) deployments has substantially expanded the attack surface of modern cyber-physical systems, making accurate and computationally feasible malware detection essential for enterprise and industrial environments. This study presents a large-scale, systematic comparison of 27 machine learning (ML) and 18 deep learning (DL) models for IoT malware detection across eight major malware categories: Trojan, Botnet, Ransomware, Rootkit, Worm, Spyware, Keylogger, and Virus. A realistic dataset was constructed using 50,000 executable samples collected from the Any.Run platform, including 8000 malware instances (1000 per class) and 42,000 benign samples. Each sample was executed in a sandbox to extract detailed static and behavioral telemetry. A targeted feature-selection pipeline reduced the feature space to 47 diagnostic features spanning static properties, behavioral indicators, process/file/registry activity, debug signals, and network telemetry, yielding a compact representation suitable for malware detection in IoT settings. Experimental results demonstrate that ensemble tree-based ML models consistently dominate performance on the engineered tabular feature set as 7 of the top 10 models are ML, with CatBoost and LightGBM achieving near-ceiling accuracy and low false-positive rates. Per-malware analysis further shows that optimal model choice depends on malware behavior. CatBoost is best for Trojan/Spyware, LightGBM for Botnet, XGBoost for Worm, Extra Trees for Rootkit, and Random Forest for Keylogger, while DL models are competitive only for specific categories, with TabNet performing best for Ransomware and FT-Transformer for Virus. In addition, an end-to-end computational time analysis across all 45 models reveals a clear efficiency advantage for boosted tree ensembles relative to most DL architectures, supporting deployment feasibility on commodity CPU hardware. Overall, the study provides actionable guidance for designing adaptive IoT malware detection frameworks, recommending gradient-boosted ensemble ML models as the primary deployment choice, with selective DL models only when category-specific gains justify additional computational cost.
The rapid adoption of real-time digital payment systems introduces cybersecurity risks that extend beyond technical vulnerabilities to include significant human and organizational factors. This study evaluates a dual-layer data protection mechanism combining AES-128 encryption with spread spectrum audio steganography within Brunei's digital payment context. Stakeholder interviews with Cyber Security Brunei (CSB), National Digital Payments Network (NDPx), and a local bank were conducted alongside experimental testing using Stripe Sandbox. Human factor issues accounted for 47% of identified cybersecurity concerns. Experimental results demonstrated that spread spectrum steganography achieved an 87.5% robustness rate across eight attack scenarios while maintaining near real-time performance with an average processing time of 568.82 ms and acceptable audio quality (PSNR 26.30 dB). Sandbox validation confirmed feasibility within realistic payment workflows. The findings support data-centric security as a compensating control in human-dominated threat environments and demonstrate the viability of combining encryption and steganography to reinforce instant payment security.
The pharmaceutical industry's R&D model faces unsustainable costs and environmental impact, yet educational pipelines lag in integrating AI and green chemistry skills essential for modern drug discovery. We conducted cheminformatics analysis of 50,000 bioactive compounds from ChEMBL v35 to quantify AI-driven discovery advantages. Natural language processing with BioBERT embeddings systematically analyzed international accreditation standards (ACPE, EAFP). AI-driven approaches achieved 99.998% reduction in experimental burden and 32% lower endocrine disruption potential versus traditional high-throughput screening. Accreditation analysis revealed a critical deficit: mean Sustainability Integration Score of only 18.3% (range: 9.5-27.1%; SD = 6.2). The observed reductions in endocrine disruption potential and predicted E-factors for AI-designed compounds should be interpreted as signals of favorable computational toxicology profiles within the dimensions assessed, rather than comprehensively validated environmental safety. We introduce the Accreditation-Aligned Sustainable Education (AASE) framework and GreenMedSim computational sandbox - an experiential platform virtualizing wet-lab activities for iterative molecular design. GreenMedSim is presented as a theoretically grounded and computationally specified educational prototype; empirical user trials and learning outcome data are planned for Phase 1 implementation (2026-2027).This work provides an actionable strategy aligning pharmaceutical workforce training with sustainable drug development, directly supporting UN Sustainable Development Goals 4.7, 9, and 12.
The low-spin electronic structure of pyrite (FeS2) severely restricts its activity in generating hydroxyl radical (•OH) through green oxidation. Herein, we report a dual-channel mechanism for in situ activation of natural FeS2 by greenhouse gas CO2 to overcome these limitations, with demonstrated potential for environmental remediation applications. Dissolution of CO2 generates H2CO3, which acidifies the aqueous environment and promotes Fe(II) release, exposing reactive surface sites. Concurrently, HCO3-/CO32- ligands coordinate with surface Fe(II) to form interfacial Fe-carbonate complexes. This coordination distorts the Fe-S lattice and triggers reconfiguration of the water-mineral interface, collectively boosting •OH generation. X-ray absorption fine-structure spectra (XAFS), Mössbauer spectra and density functional theory (DFT) calculations confirm that this reconstructed interface induces a transition of the Fe(II) center from low-spin (t2g6eg0) to intermediate-spin (t2g5eg1), narrows the bandgap, lowers the activation free-energy barrier, and enhances the electron transfer with O2, resulting in an approximately 90% increase in •OH yield. In batch experiments, degradation efficiencies of sulfamethoxazole, florfenicol, and 2,4,6-tribromophenol reached 66%, 54%, and 58%, respectively, with a significant reduction in the ecotoxicity of transformation products. Beyond batches, a two-dimensional sandbox aquifer model was conducted; after 5 days of continuous flow, degradation efficiencies were 73%, 48%, and 41%, respectively. This work advances a novel mechanism for enhanced •OH generation by CO2 and FeS2, converting low-value minerals and waste gas into effective resources for groundwater purification, thereby achieving a win-win situation of resource utilization and environmental protection.
Healthcare data exchange increasingly relies on HL7 FHIR, but FHIR's implementation complexity creates barriers for clinical workflows. Large language model (LLM) agents could bridge this gap by translating natural language requests into structured FHIR operations, yet their reliability remains unproven. We present FHIR-AgentEval, an extensible evaluation sandbox comprising 43 modular tasks for benchmarking LLM agents on realistic appointment management and genetic testing workflows. Each task executes against a resettable FHIR server with custom deterministic validation of both agent responses and resulting server state. We run an ablation study of five agent configurations, varying access to an on-demand FHIR R4 specifications server and long-term memory trained with or without specification grounding. Across four experimental settings, memory consistently improves task success and reduces strategic failures such as incorrect tool selection and resource-type confusion. On held-out tasks, the best memory configuration improves success by 9.1% over baseline, offering a potential pathway toward more robust clinical deployment.
Most studies about the groundwater circulation well (GCW) and relevant approaches coupled with other remediation techniques were conducted under ambient temperature conditions. To address this gap, using a series of laboratory-scale sandbox experiments, complemented by numerical simulations, the present study aims to evaluate the remediation performances of this coupled thermal conductive heating (TCH) - GCW system for both the conservative tracer (brilliant blue) and the semi-volatile organic pollutant (nitrobenzene). For the conservative tracer, the remediation processes of the GCW were significantly expedited by heating, but the final remediated area and the remediation efficiency demonstrated similar results for different heating conditions. For example, under the heating temperature of 110 °C, the required remediation time consumption results for observation points had reduced by 2.5%-50% compared with the no-heating condition, but the final remediated area results under heating conditions only increased by less than 6%. For nitrobenzene, higher heating temperatures can lead to better remediation performances, especially in tailing zones. Under the heating conditions, the remediated areas of nitrobenzene increased from 53.95% to 99.74%, and the required remediation time consumption results decreased by 4%-95%. Based on scenario analysis using numerical simulations, the relative contributions of four mechanism processes for removing nitrobenzene using the TCH-GCW approach were quantified and ranked as: convection > > hydrodynamic dispersion > volatilization > adsorption, while the removal of the conservative tracer was attributed to convection. The removed masses and relative contributions of nitrobenzene due to dispersion and volatilization increased as the heating temperature increased.
Internet of Things (IoT) devices are highly vulnerable to botnet attacks because they are heterogeneous, resource-constrained, and often weakly managed. Although many existing studies focus on IoT botnet detection, the post-detection stage remains comparatively underdeveloped, particularly when mitigation must be rapid, lightweight, and able to preserve service continuity. This paper presents Myco-Barrier, a bio-inspired post-detection mitigation architecture implemented as a virtual dynamic demilitarized zone (DDMZ) at the network edge. The framework translates principles from fungal mycorrhizal networks and the blood-brain barrier into an SDN-controlled overlay that supports three complementary mitigation modes: Myco-Scout for rapid isolation, Myco-Box for sandbox-based forensic routing, and Myco-Swap for continuity-preserving virtual proxy substitution. Myco-Barrier assumes that suspicious behaviour has already been identified by an upstream intrusion detection system and focuses on mitigation orchestration after alert generation. The framework is evaluated through a dual-tier methodology consisting of large-scale discrete-event simulation and Mininet/OVS emulation. The evaluation indicates that Myco-Scout provides the fastest containment and lowest control-plane actuation latency, whereas Myco-Swap provides the strongest service continuity, maintaining about 85% legitimate delivery under the evaluated flooding scenario. Overall, these findings support the feasibility of edge-orchestrated post-detection mitigation and clarify the trade-offs among containment, inspection, and service continuity in heterogeneous IoT environments.
暂无摘要（点击查看详情）
This paper presents Jub, a Life Science and Healthcare Data Platform (LSHDP) based on generic sandboxes that integrate AI tools and cloud storage into big data science services. Jub automatically and transparently creates data science services to transform datasets into massive information products by using a profiling methodology. These products are presented by generic-secure cloud-based FAIR observatories adding Programmable, Configurable/Customizing, Adaptable, and Resiliency properties (PCA-FAIR-R). This enables organizations to conduct and customize complex analytics processes to support decision-making. We conducted a study case to convert mortality, climate, and pollutants datasets (2000-2023) reported by the Mexican Government into a solid core hub of information products: 16 strategic data observatories based on 85,171,404 information products created from 114,155,622 spatio-temporal profiles of the International Classification of Diseases (ICD-10) mortality classes/strata and cancerogenic substances. An exploratory study revealed highlights about the significance of breast cancer mortality rate growth showing possible associations with air pollutants. This paper also describes the lessons learned from the practice and experience of implementing Jub sandboxes-based observatories for the Population-based Cancer Registry Network deployed on the Mexican territory in 12 Mexican states by public healthcare institutions, as well as to implement bone cancer deep-learning-based diagnosis at a national Hospital.