The expanding role of intelligent systems in biomedical science marks a shift from passive analysis towards active participation in discovery and care. Recent scholarship has begun to frame these systems not merely as models, but as agents capable of planning, interaction, and adaptation. This letter reflects on such developments, acknowledging their conceptual clarity and practical ambition, while raising questions about evaluation, responsibility, human judgement, and long-term scientific culture. The intent is to encourage careful reflection as these technologies move closer to real-world integration.

This letter comments on the recently published AICellType platform for large language model (LLM)-based cell type annotation in single-cell and spatial transcriptomics. While appreciating the authors' systematic benchmarking and practical contribution, concerns are raised regarding the continued dependence on proprietary commercial LLMs such as Claude 3.5 Sonnet for biomedical annotation. Greater emphasis is suggested on open-source biomedical LLMs, multimodal vision-language models, local deployment, reproducibility, privacy preservation, and regulatory compliance to ensure more transparent, reliable, and sustainable medical annotation systems for translational bioinformatics.

ABSTRACT The rapid expansion of generative AI—particularly large language models (LLMs)—into mission‐critical domains has underscored the urgent need for unified frameworks that embed trust, risk and security management (TRiSM) throughout the AI lifecycle. In this work, we present a comprehensive review and synthesis of AI TRiSM, uniting five foundational pillars: explainability with real‐time drift monitoring, ModelOps governance, application‐level security, data protection and privacy, and adversarial resilience. We introduce three aligned taxonomies for trust dimensions (e.g., fairness, transparency, accountability, inclusiveness, ethical alignment), risk categories (e.g., model, data, legal, operational, societal, cognitive, emergent, third‐party) and security controls (e.g., access management, infrastructure hardening, runtime enforcement, privacy‐enhancing techniques). Building on these, we develop a detailed toxicity taxonomy for generative AI—covering hate, violence, self‐harm, misinformation, bias, jailbreak attacks, multimodal harms, and more—each mapped to specific TRiSM safeguards. Through cross‐domain case studies in finance, healthcare, autonomous vehicles, public sector, cybersecurity, and beyond, we illustrate practical integration patterns and governance workflows. We also identify key adoption challenges—fragmented tooling, late‐stage governance, scalability constraints, evolving threats and regulations—and chart a forward‐looking roadmap toward adaptive, AI‐driven policy engines, causal explainability, privacy‐by‐design pipelines, continuous real‐time assurance, federated governance, quantum‐safe architectures, and sustainable “green AI” practices. This article aims to guide researchers and practitioners in designing, evaluating and scaling resilient, ethical, and compliant AI systems at enterprise scale.

ABSTRACT Neuromorphic computing is emerging as a promising paradigm for sustainable edge intelligence by enabling event‐driven, low‐latency and energy‐aware computation close to sensors. However, the field remains fragmented across device technologies, mixed‐signal and digital hardware platforms, spiking neural network models, software frameworks, event‐stream processing tools, interoperability standards and benchmarking practices. This review provides a cross‐layer synthesis of contemporary neuromorphic computing platforms with particular emphasis on their relevance to scalable and sustainable edge deployment. The article organizes the neuromorphic ecosystem into interconnected layers spanning materials and devices, hardware architectures, software and interoperability tools, benchmark resources and application domains. Mixed‐signal platforms are analysed in terms of analog efficiency, accelerated neural dynamics, biological plausibility, calibration requirements, variability and reproducibility challenges. Digital neuromorphic processors are examined with respect to programmability, deterministic execution, routing fabrics, memory organization, software integration and deployment readiness. The review further discusses software frameworks, simulators, event‐data libraries, hardware‐mapping tools and intermediate representations that support model development, portability and cross‐platform evaluation. A central finding is that neuromorphic systems cannot be compared meaningfully using isolated metrics such as neuron count, chip power, latency, or throughput alone; fair evaluation requires explicit reporting of workload, event rate, model topology, mapping strategy, software stack, measurement boundary and deployment context. Accordingly, the article proposes a benchmarking and reporting perspective for neuromorphic edge intelligence that links accuracy, latency, energy efficiency, robustness and reproducibility. Thus, this review clarifies current progress, unresolved challenges and future directions for sustainable edge sensing, robotics, healthcare monitoring, smart infrastructure, industrial automation and distributed intelligent systems.

This letter critically evaluates the feasibility of implementing open-source large language models in regulatory research, building upon the recent study on zero-shot and few-shot learning approaches for regulatory tasks. While the study demonstrates that models like Flan-T5 can effectively extract pharmacokinetic drug-drug interactions and intrinsic factors from Food and Drug Administration (US) drug labels with high precision, it also highlights significant challenges, including computational constraints, performance variability, prompt sensitivity, and the risk of misclassification. To address these issues, this letter discusses intuitive ways for mitigating these limitations.

ABSTRACT We present the first comprehensive, end‐to‐end quantitative evaluation of the linguistic relativity hypothesis in AI‐generated text, using ChatGPT‐4o mini to generate responses to 10 culturally salient prompts across 13 typologically diverse languages. Semantic shifts were quantified using pairwise cosine similarity scores computed from multilingual MiniLM sentence embeddings. A one‐way analysis of variance (ANOVA) reveals statistically significant variation in semantic alignment across language pairs, with , , and effect size . These results are further supported by a non‐parametric Kruskal–Wallis test yielding , , indicating robust differences in distribution. Prompt‐specific semantic shifts also exhibit significant variation, as shown by ANOVA results , , and . Sentiment polarity analysis using the Polyglot toolkit reveals significant effects of language on sentiment distribution, with , , and . Disaggregated analysis shows that positivity ratios differ by prompt (, , ), while negativity scores display even greater divergence across prompts with , , and . An unsupervised clustering procedure () classifies languages into three distinct groups based on semantic alignment: (i) high‐alignment (), (ii) intermediate (), and (iii) neutral‐tone clusters. Each group exhibits distinctive polarity profiles, with median sentiment polarity ranging from to . These results demonstrate that linguistic structures exert a measurable influence on AI‐generated content, underscoring the need for culturally sensitive AI design practices. These results affirm that ChatGPT‐4o mini's outputs align with the linguistic relativity hypothesis, clearly illustrating that language structures significantly shape AI‐driven interpretation All associated code and data are available in the GitHub repository: https://github.com/ParthaPRay/Liguistic_Relativity_Chatgpt .

Dear Editor, I am writing in response to the recent publication “Using AI-generated suggestions from ChatGPT to optimize clinical decision support.”1 This ground-breaking study provides an essential exploration into the potential of artificial intelligence (AI), specifically large language models, to augment the logic of clinical decision support (CDS) alerts, thus addressing a critical challenge within the healthcare industry. While this is a promising stride forward, I would like to shed light on some limitations and offer suggestions for future research. First, the research underscores the considerable potential of the ChatGPT model in understanding and providing relevant suggestions to optimize CDS. However, the model's applicability is compromised by the cut-off of its training data, which only extends up to the year 2021. This is a significant concern, as it means the model is incapable of including advancements or changes in medical knowledge that occur beyond this year. It is a common and pressing issue in the implementation of AI in healthcare—the necessity for the model to be constantly updated to remain concurrent with the dynamic nature of medical knowledge. It would be valuable for future investigations to consider the logistics and feasibility of real-time or frequent updates to the model's training data.

I am compelled to extend my profound respect and commendation for the article titled ‘Opportunities and challenges for ChatGPT and large language models (LLMs) in biomedicine and health’ by Tian et al. [1], featured in Briefings in Bioinformatics. This meticulous exploration into the intersection of LLMs, such as ChatGPT with biomedicine and healthcare, stands as a significant beacon for researchers, healthcare professionals and policymakers alike. It offers a balanced and critical examination of the transformative potential these technologies hold, alongside a candid discussion of the multifaceted challenges and ethical dilemmas they introduce [2, 3]. The authors have commendably navigated through the large landscape of LLM applications in biomedicine, illuminating areas, such as biomedical information retrieval, question answering, medical text summarization, information extraction and medical education. This breadth of exploration provides a comprehensive overview that not only underscores the innovative capabilities of LLMs to revolutionize healthcare practices but also conscientiously highlights the need for caution, especially in areas concerning data privacy, ethical considerations and the mitigation of biases. We enlist some more popular LLMs in Table 1 in addition to what the authors did in their article. A brief list of LLMs on health domain PT, PyTorch; ST, SafeTensors. A brief list of LLMs on health domain PT, PyTorch; ST, SafeTensors. The article pin points the limitations of LLMs as: hallucination, fairness and bias, privacy, legal and ethical concerns, lack of comprehensive evaluations, open-source versus closed-source LLMs and open-source versus closed-source LLMs. We add some key new challenges, such as those below along with their mitigation strategies, as shown in Table 2. Further issues and mitigation strategies by using LLMs in health domain Further issues and mitigation strategies by using LLMs in health domain The article suggests some key applications of LLMs in biomedicine and healthcare, including information retrieval, question answering, biomedical text summarization, information extraction and medical education. We include some new ways where LLMs can be beneficial in healthcare domain such as: (i) personalized treatment recommendations, (ii) predictive health analytics, (iii) automated clinical coding, (iv) virtual health assistants, (v) genomic data interpretation, (vi) patient journey mapping, (vii) healthcare workflow optimization, (viii) telemedicine support systems, (ix) ethical decision-making support, (x) clinical trial participant matching, (xi) drug discovery and repurposing, (xii) mental health monitoring and support, (xiii) nutritional advice with lifestyle coaching and (xiv) automated medical literature review. As we stand on the precipice of this AI revolution in healthcare, it is imperative that we proceed with cautious optimism, guided by a commitment to ethical principles, inclusivity and the unwavering pursuit of advancements that serve the greater good [4]. The journey ahead is fraught with challenges, but with collective wisdom, collaboration and ethical observance, the integration of LLMs into healthcare promises to open new horizons for medical science and patient care [5]. Finally, Tian et al. have provided a timely and comprehensive review of the opportunities and challenges of ChatGPT and LLMs in biomedicine and health. While the article serves as a valuable resource for researchers, healthcare practitioners and policymakers, it could have benefited from exploring some futuristic challenges and applications of LLMs in the biomedical domain. Nonetheless, their work contributes significantly to the ongoing discourse on the role of AI in healthcare and paves the way for future research and development in this field. Addressing the new challenges posed by LLMs in healthcare. Inclusion of new LLMs for healthcare. Navigating new mitigation strategies. Addition of novel LLM applications in medical domain. The author thanks Claude3 for initial discussion and drafting of this article. P.P.R. conceptualized, analysed, and drafted the manuscript. Partha Pratim Ray is working as an active academician in the field of next-generation technologies. He has published more than 140 research papers till now. He was listed as one of the top 2% scientists in the world by the Stanford University ranking in 2020, 2021, 2022 and 2023. He has a keen interest in conducting research in the key and cutting-edge technological domains. He is presently serving as an assistant professor in the Sikkim University, India. He has 7829 Google Scholar Citations, h-index of 33 and i10 index of 64. He is a senior member of IEEE and a fellow of IETE.

Internet of Things (IoT) has revolutionized the way of augmenting the physical-digital interaction. However, it lacks the holistic conceptualization about the integration notions which is related to the human-in-the-loop centric design paradigm. Thus, dew computing comes into the scenario to improve the human- aware personalized service provisioning at the edge of the network. A proper alignment between IoT and dew computing can minimize the gap of realizing the human-in-the-loop aspect in correlation to the IoT domain. This article presents a novel dew- IoT architecture to resolve such issues while inculcating dew computing-based metrics as a key enabler. We also discuss key requirements behind the integration of dew computing to IoT for efficient humanized service mitigation. We also highlight key challenges and provide the future road map.

The pervasiveness of newly introduced Internet-of-Things (IoT) devices has opened up new opportunities in healthcare systems, for example in facilitating remote patient monitoring. There are, however, security and privacy considerations in the transmission of data from these devices to the backend server, and across heterogeneous IoT networks. In this study, we propose a novel privacy-preserving scheme which is based on blockchain and swarm exchange techniques to facilitate seamless and secure transmission of user data (e.g., electronic health record (EHR)-related information) through secured swarm nodes of peer-to-peer communications. BIoTHR refers to the proposed scheme on the private blockchain-assisted EHR management using IoT. Specifically, new blockchain and swarm exchange infrastructures are suggested as a backbone of the proposed scheme to ensure secure and reliable data transmission and timely monitoring of data sent across IoT networks. An autonomous encryption-decryption mechanism is also utilized, along with a dynamic and modular server assistance technology to deploy EHR transmission in a secure manner. Moreover, several swarm-listen, announcement, peer open and peer closing algorithms are incorporated to employ the actual power of pervasive EHR transmission for better e-healthcare service provisioning. The proposed scheme is developed using the open-source tools of GnuPG, IPFS, and Golang. Proposed study simulates a number of heterogeneous IoT-based health sensor nodes, namely, body temperature, pulse rate, and oxygen saturation, i.e., SPO2, galvanic skin response, and blood glucose in blockchain-assisted swarm exchange framework. The results reveal that the proposed scheme, in terms of blockchain-IoT, swarm exchange and EHR transmission, outperforms several peer techniques.

Detrimental effect of bad oral habits, such as smoking and chewing tobacco, on chronic periodontitis (CP) manifest chronic inflammation of gingival tissues which majorly results in gum bleeding, and teeth loss. A genetic association study of Interleukin 1 beta (IL1β) has been conducted in CP patients having smoking and chewing tobacco habits in regular life style. A molecular docking study has been consequently done to analyze the effect of tobacco on CP progression in depth. All statistical evaluation has been done by using SPSS v16.0. The findings of the study show the significant association of IL1β gene polymorphisms with CP increased susceptibility in combination of oral habits as mentioned earlier. The docking profile has showed the highest binding affinity of IL1β protein with the Nicotine derived Nitrosamine Ketone (NNK), one of the derivatives of nicotine which is in-taken through the habits associated with smoking and chewing tobacco. Nicotine, N-nitrosoanabasine, and N-nitrosonornicotine, the other derivatives, have also demonstrated significant impact over the IL1β protein-caused altered expression. Thus, this study concluded that the harmful effect of tobacco may increase the inflammation in periodontia by inducing the inflammatory active site of the IL1β protein in the CP patients.